Small Packet Communication Network

1. Field of the Invention

This invention relates to communication networks and more particularly to communication networks using individually switched small blocks of data for communication through the network.

2. Discussion of the Prior Art

There exists today a great demand for rapid and efficient communication of digital data through a network. Rapid response is particularly required for such applications as interactive communication between a plurality of terminals and a host data processor as in an educational terminal system. Such a system is particularly demanding because of the nature of the interactive communication between a terminal and a host computer. A terminal keystroke may require a host response in which enough data is communicated back to the terminal to fill a CRT display. If even short, perceptible delays occur in the communication of this information, the terminal user may become discouraged and the value of the system is impaired.

While a number of data communication systems are currently known, they fail to provide the combination of reliability, economy and rapid response that is available with the present system. The most straight forward approach to data communication is to provide a dedicated communication link between each terminal and the host. The link can be provided either continuously or only upon demand while a terminal is operational. Such an arrangement solves the delay problem because the link is always available for the communication of data. However, a link that is capable of carrying high speed digital data is quite expensive and the utilization of a dedicated data link is very poor because data is actually being sent to or from an operational terminal only a small percentage of the time. Where any substantial distances are involved the costs can be prohibitive for many network applications, including educational terminals.

An approach which helps to reduce the cost of a communication link is to employ switching to permit sharing of a link. A complete communication link is established between a sender and a recipient of data, but only during the times that data is actually being transmitted. Such a system still remains quite expensive and cannot accommodate well multiple simultaneous requests for the data link during times of peak demand.

Further data link efficiencies are realized by a technique known as packet switching in which a block of data, typically having 1024 bits or more is sent as an individual entity. The blocks are sent through the network from switching point to switching point and different portions of a data link can carry different data blocks simultaneously. There is no need to establish at one time a dedicated path from sender to receiver and blocks relating to different senders and recipients can be interspersed on a shared data link segment to maximize the efficiency of use. However, substantial communication data or "overhead" is required just to direct the data blocks through the switching network and the delay between sending and receipt may be substantial from the viewpoint of an interactive terminal user. In addition, the block or packet switching sytem does not readily accommodate peak demands in which communication demand exceeds the capacity of the data links. Such a system is discussed in articles by Lawrence G. Roberts, "Data by the Packet," Spectrum pages 46-51, Vol. 11, No. 2 (February 1974 ) and by R. E. Kahn, "Resource-Sharing Computer Communication Networks," Proceedings of the IEEE, pages 1397-1407, Vol. 60, No. 11 (November 1972).

One arrangement using time division multiplexing with dedicated time slots is shown in U.S. Pat. No. 4,007,441. Small, 4 element communication rings are established with communication processor coupling the small rings to terminal devices or other rings.

Still another arrangement that is commonly known as a Pierce Ring is illustrated in U.S. Pat. No. Re. 28,811 and described in an article Pierce, J. R. Coker, C. H., and Kroppl, W. J., "Network for Block Switching of Data" IEEE Conv. Rec., New York, March 1971. Another loop arrangement is taught by Fraser, "Digital Data Communication System," U.S. Pat. No. 3,749,845. In this arrangement a series of interconnected loops of different capabilities interconnect the devices of a network. Such a system is further discussed in an article by Hayes, J. F. and D. N. Sherman, "Traffic Analysis of a Ring Switched Data Transmission System", The Bell System Technical Journal, Vol. 50, No. 9 Pages 2947-2978 (November 1971).

Further background on data communications can be obtained from Proceedings of the IEEE, Vol. 60, No. 11, November 1972, which is a special issue on computer communications.

Additional background information is available from Hayes, J. F. and D. N. Sherman, "A Study of Data Multiplexing Techniques and Delay Performance", The Bell System Technical Journal, pages 1983-2011, Vol. 51, No. 9, (November 1972) and from "Computer Network Requirements Study" prepared by Network Analysis Corporation, Beechwood, Old Tappan Road, Glen Cove, N.Y. in March 1974.

A data communication system in accordance with the invention provides rapid response interactive communication between a plurality of digital data devices. For example, the system forms a network providing communication between a plurality of terminal devices such as those which convert between machine usable and human recognizable forms of data representation and a communicating facility such as a host data processor or a trunk switching system. The system may include a host data processor with a large data base at a central processing location which is coupled for communication with other data processors, a plurality of terminals and a communication network including a network control unit controlling data flow through data channel to provide efficient data communication between the terminals and the host.

The network is particularly useful for communication involving a central processor and a large number of terminals that are separated from the host and from each other. While the present arrangement is specially adapted to accommodate a large volume of traffic flowing downward from the host and a smaller volume of higher priority traffic flowing upward to the host, the system is readily adaptable to other traffic flow patterns using the principles described herein. For example, communication may be predominantly between terminals instead of between terminals and host and may be between terminals of one network and terminals of another network, which may be coupled to a different host with trunking provided between hosts.

Each network is highly modular in nature and includes a network controller coupled between a host data processor and one or more subnetworks through fan out node devices. The network controller provides network supervision, maintains a data base of network operational parameters and separates upward and downward flowing traffic between the network and the host. Actual data flow is substantially independent of the controller with control over data flow being distributed throughout the network.

Each subnet is headed by a line master exchange unit device which contains substantial buffer data storage. Each line master exchange unit device is preferably located for high speed, parallel data communication with the host and is coupled through a serial data link to the individual terminals of the subnet. Host supplied data is stored in the large buffer storage until it can be communicated at a slower rate to buffer storage at a terminal node device or unit at each terminal site.

A line master can be coupled to two serial data loops which typically extend in parallel, opposite flow direction arrangement to provide a single full duplex loop. Such a loop is available as a 9.6 KB digital data service from telephone utilities. The duplex loops provide long distance communication as between cities and buildings within the cities while line exchange units disposed therealong provide 57.6 KB coaxial cable local simplex loops. The individual terminals are coupled along these loops through terminal node units and terminal exchange units to a communication medium such as a host data processor.

The network is capable of moving data much faster than the individual terminals can actually receive it. It thus becomes possible for the network to store a backlog of terminal data both at the subnet level and at each terminal point. A network line allocation algorithm at the line master exchange unit device monitors the availability of terminal data at the subnet level and each terminal usage rate to maintain a buffer data store at the terminal end of the network at a maximum. Thus, during short periods of peak overload, all of the terminal buffers are drawn down slowly and uniformly and each terminal continues to receive data at its normal specified rate. Only during an unusual and extended peak overload can a terminal user notice a pause in the receipt of data. This arrangement allows for higher loading of the communication channels without degradation of service.

The highly modular network is implemented with programmable micronodes and two port memories which implement data queues to provide communication between micronodes. The micronodes and two port memories are themselves implemented with repetitive modular components such as memory arrays, microprocessors, channel input modules, channel output modules and interface modules selected in accordance with a function to be performed. Costs for manufacturing, repair and spare parts storage are thus reduced.

All information, whether data or supervison is communicated through the network over the same communication facilities using a standardized small data block format. Each block contains 24, 8 bit bytes including three header bytes and a check sum byte to leave 20 bytes for actual data. Up to 16 blocks make up a message and multiple messages may be used to send unlimited data. The standard format minimizes the header data while the smaller block size and a special thread through technique assure a rapid time response. Complete data blocks are typically assembled only at buffer storage at the line master, the line exchange unit and at the destination terminal exchange units. In between, each block is handled on a double byte basis. It is possible for the line exchange unit to receive the first byte of a block before the line master sends the last byte, and for a terminal node to receive the first byte of a block before the line exchange unit sends the last byte with intermediate bytes being distributed along the respective connecting channels and nodes. Avoidance of the time delays required to accumulate a complete block at one node prior to communication to the next node greatly improves response time. A high priority is assigned to upward flowing messages to minimize the delay before the beginning of a response reaches the slower terminals.

Along the serial communication channels further efficiency is attained by transforming each 24 byte block into a single or double small block of 12 bytes each. Wasted communication capacity for less than full blocks is thus reduced. Each small block is communicated serially in one 96 bit data frame along the communication link. The data or time frames occur sequentially and small blocks are added to empty frames or pulled off as appropriate. Empty frames are unique in that they contain a logic 1 start bit followed by 95 zeros and they are used for automatic self-synchronization by exchange units connected along a serial data link.

The network provides dynamic self-determination of its own structure. Most of the micronodes have an address assigned thereto which is unique within a network. Every 22 seconds the network controller issues an all call message which reaches all addressable nodes. These nodes respond with a data block carrying a selected one of several performance parameters and the responding node address. The node directly above each responding node inserts its address into the response message to permit dynamic reconstruction of the network hierarchy. In order to distribute the all call responses and prevent a sudden overloading of the network, each node delays a time proportional to its node address before responding. Furthermore, each higher level node maintains an address table of lower level responding nodes to permit proper routing of downward flowing messages. Terminals or other components may thus be added or subtracted from the network and an adjustment of the network hierarchy and network address tables will automatically occur as a result of the next all call messsage.

An error control loop is maintained on a quasi-message basis between a line master exchange unit device and the terminal exchange units for the destination terminals beneath it. When a terminal exchange unit indicates the availability of buffer storage space for a message the line master begins transmitting a message as sequentially numbered blocks. The terminal exchange unit stores each block according to a sequence number and simply discards erroneous blocks. A pointer is maintained for sequential, error free blocks. If a block is received that is too far out of order an error is presumed and an error message is sent to the line master indicating the last complete sequential block. Retransmission resumes from that point on. An acknowledgement message is sent upon receipt of a complete message and a time out condition can generate additional error messages. This system is extremely simple and requires very little communication overhead, yet provides effective error control with very little data redundancy and rapid detection of an error.

Using the inherent system communication facilities, any addressable location at any addressable micronode can be read from or written into, if that is possible, i.e. writeable memory. The entire system can thus be selectively programmed under control of the host or network controller for operation as a giant multiprocessor. During operation as a communication network, the exact status of each node can be checked for diagnostic purposes.

The line master exchange unit periodically checks for continuity on the duplex line connected thereto. If a break is found, it automatically operates to reconfigure the duplex line as two separate simplex loops. Any break is isolated in such a way that communication is maintained with a maximum number of terminals while the break is being repaired.

The network moves the logical interface to the terminal to a node located at the terminal site. This allows sharing of communication channels at all levels in the network, including not only trunking channels (as with current packet switching networks) but also sharing of channels used for local distribution.

Responsiveness in a network increases with data link speed and decreases with packet length. By using efficiently blocks an order of magnitude shorter than current practice, the network structure provides the necessary responsiveness for good man/machine interaction over the low speed data links used for local distribution.

A better understanding of the invention may be had from a consideration of the following Detailed Description taken in conjunction with the accompanying drawings, in which:

FIG. 1 is a block diagram representation of a communication system in accordance with the invention;

FIG. 2 is a block diagram representation of a host adapter used in the system shown in FIG. 1;

FIG. 3 is a block diagram representation of a network controller used in the system shown in FIG. 1;

FIG. 4A is a block diagram representation of a line-master exchange unit device used in the system shown in FIG. 1;

FIG. 4B is a waveform diagram which is useful in understanding the operation of the linemaster exchange unit device shown in FIG. 4A;

FIG. 5A is a block diagram representation of a 9.6KB line exchange unit device used in the system shown in FIG. 1;

FIG. 5B is a waveform diagram that is useful in understanding the operation of the line exchange unit device shown in FIG. 5A;

FIG. 6 is a block diagram representation of a terminal exchange unit device used in the system shown in FIG. 1;

FIG. 7 is a block diagram and schematic representation of a small two port adapter module used in the system shown in FIG. 1;

FIG. 8 is a block diagram and schematic representation of a microprocessor unit module used in the system shown in FIG. 1;

FIG. 9 is a block diagram representation of a node address module used in the system shown in FIG. 1; message SETAKM, channel the then

FIG. 10 is a block diagram and schematic representation of a bus extender module used in the system shown in FIG. 1;

FIG. 11 is a block diagram and schematic representation of an exchange unit control module used in the system shown in FIG. 1;

FIG. 12 is a block diagram and schematic representation of a console interface module used in the system shown in FIG. 1;

FIG. 13 is a schematic representation of a timer interrupt module used in the system shown in FIG. 1;

FIG. 14A is a schematic and block diagram representation of a coax input module portion of a coax transceiver module used in the system shown in FIG. 1;

FIG. 14B is a schematic representation of a coax output module portion of a coax transceiver module used in the system shown in FIG. 1;

FIG. 15 is a block diagram and schematic representation of a phase converter module used in the system shown in FIG. 1;

FIG. 16A is a block diagram and schematic representation of a channel input/output module used in the system shown in FIG. 1;

FIG. 16B is a waveform diagram of signals that are useful in understanding the operation of the channel input/output module shown in FIG. 16A;

FIG. 17 is a block diagram and schematic representation of a channel output module used in the system shown in FIG. 1;

FIG. 18 is a block diagram representation of an 8K.times. 8 RAM array module used in the system shown in FIG. 1; and

FIG. 19 is a schematic and block diagram representation of a large two port adapter module used in the system shown in FIG. 1.

I. general Configuration

A. host Adapter 32

B. small Two Port Buffer Memories

C. queue Implementation the

D. network Controller 50

E. fan Out Node Device 54

F. line Master Exchange Unit Device 60

G. line Exchange Unit Device

H. terminal Exchange Unit Device

Ii. communication Network Modules

A. small Two Port Adapter Module 340

B. ram array Module 360

C. mpu module 370

D. node Address Module 390

E. bus Extender Module 400

F. exchange Unit Control Module 414

G. programmable Read Only Memory

H. net Master Interrupt Module

I. console Inerface Module 430

J. terminal Interface Module 446

K. coaxial Cable 57.6KB Transceiver 454

L. clock Multiply Module

M. phase Converter Module 472

N. loopback Module

O. channel Input/Output Module (CHIOM) 490

P. channel Output Module (CHOM) 540

Q. input Matching Module (IM96M) and Output Matching Module (OM76M)

R. channel Input Module (CHIM)

S. 8k.times.8 ram array Module (RAM[8K]) 560

T. large Two Port Adapter Module (L2PADAPT) 576

U. processor Input Interface Module (PIINTM) 100 and Processor Output Interface Module (POINTM) 102

Iii. message Formats

A. general

B. abort Message ABORTM) Output 32

C. input Data Message (INDATM)

D. output Data Message (ODATM)

E. permission To Send Message (PRSNDM) Input 36

F. blockset Acknowledge and Permission To Send Message (ACKPRM) Input 32

G. sequence Alarm or Status Request Response Message (ALMSRM) Input 33

H. request Status Message (REQSTM) Output 40

I. blockset Acknowledgement Message (SETAKM) Input 34

J. reconfiguration Messages

K. loopback Message (LPBAKM) Output 38

L. cancel Loopback Message (CLPBKM) Output 34

M. loop Continuity Message (LPCNTM) 35

N. network Control Messages

O. network Control Read Message (NCRDM) Output 33

P. network Control Response Message (NCRSPM) Input 37

Q. network Control Write Message (NCWRM) Output 35

R. all Call Messages (ALCALM) Output 0-5

S. all Call Response Message (ALCLRM) Input 0-5

T. reinitialization Message (REINM) Output 41

U. net Control Interface Messages

V. all Hierarchy Pair Response Message (AHPRM)

Iv. programming

A. general

B. processor Input Matching Node

C. processor Output Matching Node

D. net Master Interface Node

E. net Master Node

F. console Interface Node

G. y node

H. fan Out Node Devices

I. node B96BN

J. line Master Node

K. output End Nodes

L. input End Nodes

M. drop Nodes

N. coax Output End Node

O. coax Input End Node

P. terminal Node

Q. terminal Matching Node

V. network Node Program Address Map

Vi. network Node Program Load Lists

Referring now to FIG. 1, a digital data communication system 10 in accordance with the invention includes a plurality of host data processing systems 12, 14, 16 and 28 which are interconnected by conventional data communication links 20. While a particular pattern of data communication links 20 is provided between the various host data processors, by way of example, the exact pattern is not material to this invention and in general only a single host data processor such as host 12 may be present in the system. The host data processors 12, 14,16 and 18 may be any suitable commercially available data processor system which maintains a data base of information and makes portions of the data base available through a communication network to other devices upon demand.

Host data processor 14 is shown as being coupled through a host adapter 22 to a communication network 24 which provides communication between the host 14 and a plurality of terminal devices illustratively represented by terminal devices 26, 28 and 30. 12, 14, 16

Host data processing system 12 is shown as being coupled through two host adapters 32, 34 to two communication networks 36, 38 respectively. Network 36 provides communication between host 12 and a plurality of terminal devices 40, 41, 42 and 43 while communication network 38 provides communication between host 12 and a plurality of terminal devices 46 47 and 48. Network control may be provided through one of the terminal devices which is connected to the network in the same manner as any other terminal device and thus terminal device 43 has been designated a network control terminal.

While the exact configuration of the particular communication networks 24, 36 and 38 will depend upon the numbers and physical locations of the terminal devices connected thereto, the various communication networks may be constructed in accordance with the same principles using identical types of modular communication elements and communication data links which are illustrated. For this reason, only the particular data communication network 36 will be described in detail and it will be understood that the communication networks 24 and 38 may be constructed in a similar manner.

The network 36 includes a network controller 50 which is connected to a TTY control console 52 and which is also coupled between the host adapter 32 and a fan out node device 54. Fan out node device 54 is connected to one or more additional fan out node devices 56 (only one being shown for simplicity). Similarly, fan out node device 56 is connected to one or more communication subsystems as representatively illustrated by the 9.6 kilobit subsystem 58 including line master exchange unit device 60, and 9.6 kilobit line exchange unit devices 62, 63 and 64. Line exchange unit device 64 is in turn coupled to a 57.6 kilobit coaxial cable communication subsystem 70 including terminal exchange unit devices 72, 73, 74 and 75 which are in turn coupled through two port buffer storage memories 78, 79, 90 and 81 to terminal devices 40, 41, 42 and 43 respectively.

Communication between the various elements of the communication network 36 is provided by two port buffer memories similar to two port buffer storage memories 78-81. A two port buffer storage memory 84 couples network controller 50 to fan out device 54 while another two port buffer storage memory 86 couples fan out device 54 to fan out node device 56 and two port buffer storage memory 88 couples fan out node device 56 to line master exchange unit device 60. In an alternative small system network arrangement the fan out nodes 54, 56 may be eliminated with two port 88 being directly coupled to network controller 50 via data path 90 which is shown as a dashed line.

The two port buffer storage memories are used extensively throughout the communication network 36 between major elements of the network and also between programmable micronodes within the major elements of the network, each of which operates as a self-contained programmable processor to perform a designated network function. The two port buffer storage memories are random access memories with two ports, each of which controls access to the memory on different alternate memory cycles. During a cycle in which a given port is active, a device coupled to the memory has complete access to read or write into any location in the memory. During the next or alternate memory cycle a device coupled to the other port has complete access to read or write into any location in the memory. The two port memories are customarily used to implement a pair of queues by designating certain locations within a memory for the storage of network message information which is passing from the host toward a terminal and certain other locations in the memory for storage of network message data which is passing toward the host. The two port buffer storage memories are thus represented as a pair of oppositely directed arrows to represent the bidirectional message data flow provided by the pair of queues in a two port memory. In addition to message data, two port memories may store network control information for which it is desired that two different programmable micronodes or network elements both have access. For example, two port buffer storage memory data content status pointers are also stored in the two port memories and a status pointer for a first private micronode memory may be stored in a connected two port to enable a second micronode coupled to the two port memory to easily ascertain the status (e.g. empty or full) of the first micronode memory.

The communication network 36 is highly modular and submodular in nature to permit a highly flexible network configuration to be constructed of standard components. Because of the standardization of components, production cost may be reduced because of high production volumes and problems of storing, maintaining and accounting for hundreds of different subassemblies are greatly reduced. As previously discussed in connection with FIG. 1, the major modules include two port buffer storage memories, a network controller, fan out node devices, line master exchange unit devices, 9.6 kilobit line exchange unit devices, and terminal exchange unit devices. Each of these primary modules is in turn comprised of a selected combination of about 22 standardized submodules.

For convenient reference, the submodules are listed as follows:

1. S2PADAPT--small two port adapter module.

2. MPUMOD--microprocessor unit module.

3. NODEADR--node address module.

4. BUSXTEND--bus extender module.

5. EUCONTRL--exchange unit control module.

6. PROM [4K, 3K, 2.5K, 2K, 1.5K, 1K, 0.5K]--programmable read only memory with the number of 8 bit words being indicated in brackets, K being 1024.

7. RAMA [2K, 1.5K, 1K, 0.5K, 0.25K] random access memory array with the number of 8 bit words indicated in brackets.

8. NMIM--network master interrupt module.

9. CNSIFM--console interface module.

10. TINTM--terminal interface module.

11. COAXXCVR--57.6 kilobit coaxial cable transceiver module.

12. CLMX6--clock multiply module.

13. PHCM--phase converter module.

14. LOOPBACK--loop back module.

15. CHIO--channel input/output module.

16. CHOM--channel output module.

17. 9.6KX/R--9.6 kilobit RS232-C transmitter/receiver.

18. CHIM--channel input module.

19. 8KRAM--8K by 8 random access memory array.

20. L2PTADAPT--large two port adapter module used only in line master exchange unit devices.

21. Channel input adapter.

22. Channel output adapter.

The various submodules and their abbreviated mneumonics are provided at this point for convenient reference. To the extent that their construction is not readily apparent to a person of ordinary skill in the art, they will be described in greater detail below.

Referring now to FIG. 2, host adapter 32 includes processor input interface module (PIINTM) 100, processor output interface module (POINTM) 102, micronode PIMN 104, micronode POMN 106, micronode NMIFN 108, two port buffer memory 114 providing network control queue (NTCQ) 115 connecting micronode POMN 106 to micronode MNIFN 108, two port buffer memory 118 providing network transfer return queue (NTRQ) 119 connecting micronode NMIFN 108 to micronode PIMN 104 and two port buffer memory 116 providing host data down queue (HDADQ) 117 connecting micronode POMN 106 to micronode YN 110 in network controller 50. Processor interface modules 100, 102 are merely interface units connecting micronode PIMN 104 and micronode POMN 106 to input 115 output channels respectively of a host I/O channel. They merely provide a data format transformation between the network format and host I/O processor format. Their construction is conventional but depends upon the particular host I/O processor and has therefore not been shown in detail.

Processor input micronode 104 includes the standard modules MPUMOD designated PIMNMPU, PROM[3K] designated PINPROM and RAMA[0.25K] designated PINRAM. The three micronodes 104, 106 and 108 in the host adapter 32 are considered to form an interface between the communication network 36 and host 12 and hence are not micronodes which are addressable as part of the communication network 36. However, in general they could of course be assigned a network address and be addressable if desired. Processor input micronode 104 operates to communicate the network control response messages (NCRSPM) through processor input interface module 100. These messages are received through network response queue 119. Micronode 104 also communicates to processor input interface module 100 the input data and permission to send messages, INDATM and PRSNDM, received from a small two port buffer memory 112 which implements a host data up queue 113. Micronode 104 receives (and discards) message ALCLRM from 112. These messages are different types of data and supervisory or control messages which are communicated through the network. In general micronode 104 receives data and data control messages through two port 112 and supervisory and network control response messages through two port 118. Micronode 104 communicates with the network in only a 24 byte large block message format and operates in conjunction with PIINTM 100 to make a format transformation for communication with the host I/O processor.

Peripheral processor output micronode 106 receives data from the processor output interface module 102 and communicates it either through small two port buffer memory 116 which implements host data down queue 117 to Y node 110 or through small two port buffer memory 114 which implements network control queue 115 to net master interface node 108 or both. The messages flowing through HDADQ 117 to Y node 110 include the abort and output data messages, ABORTM and OTDATM, while messages flowing through NTCQ 115 to NMIFN 108 include NCRDM, NCWRM. Micronode 106 communicates with the network in a 24 byte block format and operates in conjunction with POINTM 102 to transform the format for compatibility with the host I/O processor. The standard modules which comprise POMN 106 include MPUMOD designated POMNMPU, PROM[3.5K] designated POMNPROM and RAMA[0.25K] designated POMNRAM. For some purposes POINTM 102 is considered part of micronode POMN 106.

Net master interface node 108 includes as standard modules an MPUMOD designated NMIFNMPU, PROM[3K] designated NMIPROM, and RAMA[0.25K] designated NMIRAM. It communicates data in a 24 byte large block format and, as explained above, it receives data through two port buffer memory 114 which provides the network control queue 115 from micronode 106 and provides information through two port buffer memory 118 which provides a network response queue 119 to micronode 104. Micronode 108 also communicates through two port buffer memory 130 which implements host control up queue 131 and host control down queue 132 to provide bidirectional communication between micronode 108 and net master micronode 134 (FIG. 3). Micronode 108 provides through HCUQ 131 messages DSDBM, NCRDM and NCWRM. It receives through HCDQ 132 messages NSPRM, SNSPRM, SNASRM, ASNPRM, NASPRM, ASPRM, NHPRM, SNHPRM, AHPRM, AHSPRM, NHCHRM, and NCRSPM. In general, net master interface node 108 serves as a communication link for bidirectional supervisory and control messages between net master node 134 and the host data processing system 12.

The small two port buffer memories which implement the queues provide buffered communication of data between two micronodes. Each memory includes a small two port adapter module (S2PADAPT) and a random access memory module RAMA[XK]. Either a single unidirectional queue or two queues providing bidirectional communication may be implemented by a two port buffer memory. The small two port adapter module causes the random access memory to be address accessible to each of the two micronodes connected thereto on alternate memory cycles. During the proper memory cycles all addressable memory locations are available to a micronode connected thereto.

Whether a private queue within a single micronode or a two port queue connected between two micronodes, each queue is implemented by four pointers stored at predetermined locations within a memory and a map table. While the specific queue implementation technique is not critical to network operation, it does provide an advantageous combination of simplicity and flexibility.

Each queue is divided into a number of cells ranging from 2 to 128 cells depending upon the size of the queue that is desired. Each cell in turn stores either 12 or 24 eight bit bytes of data (1 block) depending upon whether data is processed in large blocks or small blocks at the point in the network where the queue is implemented.

The four pointers are art of a 6 byte queue control package with predetermined sequential addresses which are available to the programs of the accessing micronodes. The first or smallest address location stores the most significant byte of a two byte next read pointer which defines an address in the map table corresponding to the cell into which data is to be read from next. The second address location in the control package stores the least significant byte of the map address. Because the map table stores a two byte address for each cell, the least significant bit of the second pointer byte is always 0 to point to the most significant of the two cell address bytes in the map table. The next one to seven bits define the cell number of the cell from which reading is to next occur. A queue may thus contain between 2 and 128 cells. Any remaining bits in the second byte define a portion of the address of the map table.

The map table is required to begin at an address that can be defined by merely clearing the cell number portion of the next read (NR) and next write (NR) pointers.

This arrangement allows the next read and next write pointers to be changed with a single word write access of the memory. The single word change of queue content (typically adding or deleting a single block) resolves contention problems between programs in different processors or at different priorities in the same processor in adding and deleting blocks from the same queue.

The third and fourth bytes of the queue control package define the most and least significant bytes respectively of the next write pointer. The next write pointer is implemented in the same manner as the next pointer except that it defines the address in the map table of two bytes which in turn define the beginning address of the next cell into which information is to be written.

The fifth byte of the queue control package stores the last cell pointer (LC), which defines the least significant byte of the next read pointer or next write pointer for the highest numbered cell. Similarly, the first cell pointer (FC), occupies the sixth byte of the queue control package and defines the least significant byte of the next read or next write pointer for the lowest number of 0 cell, which is the first cell of a queue.

The map table stores a two byte address for each cell in a queue. Each two byte address defines the address of the first byte of a cell corresponding thereto. Additional byte locations within a cell are accessed by incrementing the address of the first byte.

A queue is accessed in a circular manner. A microprocessor that is writing information into the queue first reads the next write pointer, which will initially point to the first cell, cell 0. It uses the next write pointer to address the map table to obtain the beginning address of cell 0. This cell 0 address is sequentially incremented each time a new byte is written into cell 0 until all of the data for cell 0 has been written. The microprocessor then returns to the next write pointer and uses the map table to obtain the beginning address for the second cell, cell 1. This process continues with the least significant byte of the next write pointer being compared to the last cell pointer each time the next write pointer is accessed. If the next write and last cell pointers are unequal, the next write pointer is incremented. As the next write pointer is incremented to point to the last cell it will equal the last cell pointer. Then, instead of being incremented, the next write pointer is made equal to the first cell pointer.

Reading from the queue is accomplished in the same circular manner except that the next read pointer is used to define (through the map table) the beginning of the next cell at which reading is to occur. As a new next read pointer is obtained, it is normally compared to the next write pointer. If they are equal the queue is deemed to be empty. In addition, when writing, the next write pointer can be compared to the next read pointer to prevent overlapping or overflow of the queue. However, in most locations in the network, overflow is sufficiently unlikely that the comparison is not made. If an overflow does occur, all of the contents of the queue are lost because the queue is considered to be empty when the next read pointer equals the next write poiner.

Referring now to FIG. 3, network controller 50 (NMEU) includes Y node YN 110, netmaster node NMN 134 which communicates through two port buffer memory 130 and two port buffer memory 136 as previously discussed, and console interface node CIFN 148. Net master 134 also communicates through two port buffer memory 144 which implements local control down queue 145 and local control up queue 146 to controller interface node 148. The messages communicated through local control down queue 145 include NSPRM, SNSPRM, SNASRM, ASNPRM, NASPRM, ASPRM, NHPRM, SNHPRM, AHPRM, AHSPRM, NHCHRM, and NCRSPM while messages communicated through local control up queue 146 include DSDBM, NCRDM, and NCWRM. The net master node 134 includes an MPUMOD designated NMMPU, a BUSXTEND module designated NMBUSX, a RAMA[2K] designated NMNRAMLO, a second RAMA[2K] designated NMNRAMHI and PROM[2K] designated NMNPROM.

Net master node 134 is the master control node for network 36. It provides the network 36 with a supervisory interface through host adapter 32 to host 12 via HCDQ 132 and HCUQ 131 and to local console 52 via two port 144 and console interface node 148. Net master node 134 relays network control messages received through HCUQ 131 and LCUQ 146 to the network via NMONQ 138. Net master node 134 manages a network wide network hierarchy data base and maintains a statistics data base of six parameters for each active node in the network 36. The statistics data base is organized according to node address.

Net master node 134 also operates to generate a network all call message ALCALM every 22 seconds which is communicated through NMOMQ 138 for duplication and distribution to each micronode in the network 36. There are six types of all call messages which are generated in sequence. Each of the six all call messages is essentially the same except that it corresponds to a different one of six parameters which is stored in the statistics table by the net master node 134. Upon receipt of an all call message, each addressable micronode within the network 36 waits for a predetermined period of time which is proportional to the address of the micronode and then sends to the net master node 134 an all call response message which includes the address of the responding micronode and one of six statistical parameters which depends upon the type of all call message which was received. The proportional delay distributes the all call response messages throughout the interval between two successive all call messages and thus avoids the problem of overloading the communication network 36 with many simultaneous responses from all of the addressable nodes in the network. The all call responses not only permit the maintenance of statistics tables for network control functions, but also permit dynamic connection and disconnection of network terminals automatically by the network 36. On the way down, the all call messages are duplicated and distributed to each possible downward path so that they arrive at all nodes in the network. Then, as the all call response messages pass upward through the network 36, each node maintains in memory a table of responding node addresses and corresponding paths over which the responding addresses are received. Subsequently, upon receipt of a downward passing message for one of the previously responding node addresses, the table is utilized to determine the proper path over which the message should be forwarded. Net master node 134 provides statistics and a hierarchy data base on demand to the host data processor 12 or through the console interface node 148 to the TTY control console 52. It also relays network node control information on demand from console 52 or host adapter 32 to the network.

Y node 110 includes an MPUMOD designated YNMPU, an RAMA[1.5K] designated YNRAM, and PROM[1K] designated YNPROM. Y node 110 receives information from peripheral output micronode 106 through two port 116 as explained previously and communicates information to peripheral processor input micronode 104 through two port 112 as previously explained. In addition, Y node 110 communicates through the fan-out node devices 54, 56 and their connecting two port memories 84, 86 and through two port memory 88 to line master exchange unit device 60. The messages which are passed downward toward two port 88 (see FIG. 1) include OTDATM, ALCALM, ABORTM, NCRDM and NCWRM, while the messages which are passed upward from two port 88 include INDATM, ALCLRM, PRSNDM, and NCRSPM. Y node 110 also communicates bidirectionally with net master node 134 through two port buffer memory 136 having net master node input queue (NMINQ) 137 and netmaster node output queue (NMONQ) 138. Messages PRSNDM, ALCLRM, INDATM, and NCRSPM are sent to net master node input queue 137 while messages ALCALM, NCRDM, and NCWRM are received from net master node output queue 138.

Y node 110 operates to move messages in a 24 a byte large block format between the net master node 134 (FIG. 3) and between the host adapter 32 and two port memory 84 for fan-out node device 54. Y node 110 duplicates and routes a copy of each upward flowing message from two port 84 through host data up queue 113 of two port 112 and through NMINQ 137 of two port 136 to net master node 134. Y node 110 also serves to merge messages received from net master node 134 through net master node output queue 138 with upward messages from two port 84.

The console interface node 148 communicates bidirectionally through two port buffer memory 144 with net master node 134 as previously explained. It provides an interface between the data format of the TTY control console 52 and the 24 byte large block data format of the network 36.

Network controller 50 also includes an exchange unit control (EUCONTRL) standard module designated NMCONTRL and a net master interrupt module (NMIM) designated NMIRQM. NMIRQM is a programmable timer that is assigned an address on an internal bus of net master node 134. It is reset by addressing the NMIRQM module and writing therein the four least significant bits of the 8 bit data portion of the bus of net master node 134. NMIRQM then times out after a period of time equal to 32.768 MSEC times the number written therein for resetting and generates an interrupt request signal IRQ which is communicated to the microprocessor in the microprocessor module, NMNMPU. NMCONTRL provides network controller 50 with 2 phase 1MHz master clock signals, run/halt control signals to facilitate program debugging, a power on signal and a master reset signal.

Two port buffer memory 136 includes an S2PADAPT designated HDA2PORT and a RAMA[0.25K] designted HDAQRAM. Two port buffer memory 130 includes an S2PADAPT designated HCD2PORT and a RAMA[0.25K] designated HCDQRAM. Two port buffer memory 144 includes an S2PADAPT designated LC2PORT and a RAMA[0.25K] designated LCQRAM.

Two port buffer memory 112 includes a S2PADAPT designated HDA2PORT and a RAMA[0.25K] designated HDAQRAM. It provides communication from Y node 110 to micronode PIMN 104. Similarly, two port buffer memory 130 includes a S2PADAPT designated HCD2PORT and a RAMA[0.25K] designated HCDQRAM. It provides bidirectional communication between net master node 134 and net master interface node 108.

Two port buffer memory 84 includes fan out down queue 1 150 and fan out up queue 1 151 to provide communication between Y node 110 and fan out node device 56. FODQ1 150 carries messages OTDATM, ALCALM, ABORTM, NCRDM, and NCWRM while FOUQ1 151 carries messages INDATM, ALCLRM, PRSNDM, and NCRSPM. Two port buffer memory 84 is assembled from an S2PADAPT designated F012PORT and a RAMA[0.25K] designated F01QRAM.

Referring now to FIG. 1, fan out node device 54 in-includes a sngle nonaddressable FON1 type micronode having MPUMOD designated FOW1MPU, RAMA[0.5K] designated FON1RAM, a PROM[3K] designated F01PROM, a NODEADR designated FON1ADR and EUCONTR designated FON1CONTRL. It duplicates and fans out the all call message, ALCALM, to four fan outs. It also maintains by all call responding node address a table indicating the one of four paths over which an all call response is received. Downward flowing messages can thus be distributed over the proper fan out path by using the table. Although not implemented in the present embodiment, the network may be expanded by using the last or 24th byte of a data block as a check sum below the fan out nodes and as a path address above the fan out nodes. Each fan out node would then use a different permutation of bits in the last byte to indicate the data path. A full 1K address set would then be available for each of up to 256 data paths. With the fan out nodes located at the host site, communication with the host can be deemed error free without a check sum.

Two port 86 includes fan out down queue 2 154 and fan out up queue 2 155 and is identical to two port 84, except that the RAMA[0.25K] is designated FO2QRAM and the S2PADAPT is designated FO22PORT.

Fan out node device 56 is identical to fan out node device 54 except that the numeral 2 is substituted for numeral 1 in the module designations.

Two port 88 includes subnet down queue 158 and subnet up queue 159 and provides a communication between fan out node 56 and line master exchange unit device 60. Two port 88 includes a RAMA[1K] designated BY2PORT and an S2PTADAPT designated BY2PRAM.

Referring now to FIG. 4, the line master exchange unit device 60 includes micronode B96BN 170 which provides bidirectional communication between two port buffer memory 88 and a large two port buffer memory 172, wich is in turn coupled to line master 9.6 KHz node LM96N 174. A line master exchange unit control module 176 provides general control signals for the line master exchange unit device 60 including a 1 MHz clock signal, run/halt control to facilitate program debugging, a power up signal indicating that the power supply is providing adequate power, and a master reset signal.

Micronode 174 communicates through a two port buffer memory 178 which defines an end up queue east EUQE 179 with a micronode IE96NE 180, through a two port 182 which defines an end down queue west EDQW 183 with a micronode OE96NW 184, through a two port 186 which defines an end up queue west EUQW 187 with micronode IE96NW 188, and through two port 190 which defines end down queue east EDQE 191 with micronode 0E96NE 192.

A standard utility supplied 9.6 KHz full duplex data link includes a pair of opposite direction 9.6 kilobit data lines 196, 198, a first data service unit 200 and a second data service unit 202. The end of line 198 which connects to data service unit 200 provides data through input matching module IM96M 204 and channel input module CHIM 206 to node IE96NE 180 and the opposite end, which connects to data service unit 202, receives data from node OE96NE 192 through channel output module CHOM 214 and output matching module OM96M 211. Similarly, the end of line 196 which connects to data service unit 200 is coupled to receive data from micronode OE96NW 184 through CHOM 208 and OM96M 205 and the opposite end which connects to data service unit 202 is coupled to provide data through IM96M 210 and CHIM 212 to micronode IE96NW 188.

The transmission and receipt of data over line 196 and over line 198 is essentially identical except for the opposite directions involved. For this reason, the communication will be described in detail only with reference to the line 196.

As shown in FIG. 4B, the data service unit 200 generates a 9.6 KHz squarewave clock signal C3A which is inverted by output matching module OM96M 205 and communicated to channel output module CHOM 208. CHOM 208 receives parallel data from micronode OE96NW and outputs this data serially as signal DAOUT synchronously with C3A. Signal DAOUT is complemented by OM96M 205 before it is provided to DSU 200.

The complemented data is carried by the communication link in complemented form to be output by DSU 202 as signal DAIN synchronously with a clock signal C1A. Input matching module IM96M 210 communicates both DAIN and C1A to CHIM 212 in their uncomplemented form where CHIM 212 uses the rising edge of C1A at the center of each bit time interval to sample signal DAIN by loading it into a serial-in-parallel out shift register.

Channel input module 212 provides data block synchronization of the serial data stream and also provides a serial-to-parallel conversion. Upon accumulation of each set of sixteen data bits or two bytes, the channel input module 212 generates an interrupt signal, IRQ, which causes the connected micronode IE96NW 188 to address a first I/O read select signal to generate a read shift register signal READSRS which causes channel input module 212 to place the contents of its shift registers onto an eight bit data bus designated DBO-7 which connects to an eight bit data bus of an MPU module of node 188. MPU module address line A0 is connected to channel input module 212 and commands a first byte of data to be placed on the data lines when at logic 0 and a second byte of data to be placed on the data lines when at logic 1.

A second addressable I/O read select output of the MPU module of micronode 188 commands a signal READCNTR which causes the contents of a double byte counter within the synchronizing circuitry of channel input module 212 to be placed on the data bus lines DBO-2. Micronode 188 is thus able at any time to determine the number of double bytes that have been received by channel input module for a given small data block. It will be recalled that information is transmitted over the communication links in the form of small data blocks having twelve bytes equal six double bytes equal 96 bits.

In a somewhat similar fashion, output end node OE96NW 184 provides in parallel byte form over data bus lines DBO-7 data information that is to be communicated by data service unit 200. A channel output module 208 receives the data from micronode 184 in parallel eight bit byte form, provides a parallel to serial conversion and outputs the data in serial form as signal DATAOUT in response to clock signal, C3A received from output matching module 205. The parallel data is provided to channel output module 208 by micronode 184 in a double byte form by addressing a first I/O write select output to activate a signal load LOADBYT1 which causes data to be transferred from the eight bit parallel data bus lines to a first eight bit shift register and by subsequently addressing a second location to activate a second I/O write select signal to generate a signal LOADBYT2 which causes data appearing on the data lines to be written into a second eight bit latch which is in turn coupled to a second eight bit shift register that is coupled in series with the first eight bit shift register. Upon loading the contents of the latches into the shift registers, channel output module 208 generates an interrupt signal, IRQ, to which micronode 184 may respond by loading another pair of bytes into the latches of channel output module 208.

Micronode B96BN 170 includes an MPUMOD standard module designated BNMPU, and NODEADR module signated BNADR, a RAMA[0.25K] module designated BNRAM , a PROM[2.5K] module designated BNPROM1, and a PROM[3.5K] module designated BNPROM2. Micronode B96BN 170 receives downward flowing information through two port 88 and stores the information in one of a plurality of queues formed within large two port 172.

Large two port 172 contains 32K bytes of random access storage and implements 32 buffer pairs 220 designated MBPO-31, with each buffer of a buffer pair designated either, 0 or, 1 and containing storage for sixteen blocks or cells of 24 data bytes each. Each connected terminal node is assigned one of the buffer pairs as the terminal nodes respond to all call messages and micronode 170 stores data coming from two port 88 in one of the buffer pairs according to the destination address indicated by the data. Micronode 170 also tests the availability of a buffer pair and indicates the availability to the host data processor 12 so that host 12 can output information for a given terminal only when space is available in a buffer pair and thus prevent a data overflow. Micronode 170 also stores in a down portion 227 of a master queue 226 implemented by large two port 172, nondata information that is flowing downward from two port 88. Node 170 also receives from a master up queue 228 implemented by large two port 172 all information that is flowing upward toward two port 88 from micronode LM96N 174. The messages handled by node 170 include messages INDATM, PRSNDM, OTDATM, NCRDM, NCWRM, NCRSPM, ALCALM, ALCLRM, and ABORTM. Micronode 170 processes messages ALCLRM, ALCALM, OTDATM, NCRDM, NCWRM and ABORTM with other messages being simply relayed therethrough.

Line master exchange unit device 60 also includes line master 96 node 174, an MPUMOD designated LMNMPU, a BUSXTEND module designated LMBUSX, a NODEADR module designated LMNADR, a ROMA[ 0.5K] module designated LMNRAM, a PROM[3K] module designated LMNPROM1, a PROM[3K] module designated LMNPROM2 , and a PROM[4K] module designated LMNPROM3.

Micronodes OE96NE 192 and OE96NW 184 each maintain a queue of small 12 byte blocks in their private stores waiting output on data links 198 and 196 respectively. The control pointers for the queue in OE96NE 192 are maintained in two port 190 and the control pointers for the queue in OE6NW 184 are maintained in two port 182. Thus micronode LM96N 174 may examine both sets of pointers and is able to determine the number of small data blocks in storage in the two queues in OE96NE and OE96NW.

In order to maintain maximum loading of the two output data links it is necessary that the two private queues in OE96NE and OE96NW contain blocks waiting for output on the two data links if any blocks are available from the buffer pairs 220 or MDQ 228 in L2PORT 172. Conversely, a large number of blocks waiting in the queues in OE96NE 192 and OE96NW 184 will increase the time required to output a newly arrived higher priority block.

For this reason, micronode LM96N 174 examines the number of blocks stored in the private queues in OE96NE 192 and OE96NW (using its access to the two port) and delivers a large 24 byte block (which may transform into either one or two small 12 byte blocks) to two port queue PDQE 191 only if the private queue in OE96NE 192 contains less than three small 12 byte blocks or to two port queue EDQW 183 only if the private queue in OE96NW 184 contains less than three small 12 byte blocks. Micronode LM96N outputs along either of the two output paths (not having more than two small blocks in the corresponding private queue) on a chance basis, depending on the order of examination.

In contrast to this chance path selection technique, the higher priority upward flowing messages are assigned to flow through a shortest path as explained below.

Referring now to FIG. 4, line master micronode 174 executes an availability algorithm to send information from a master buffer pair 220 to a matching buffer pair in Terminal Exchange Unit Device (40, 41, 42 or 43) when space is available within the destination buffer pair. Node 174 maintains a table of information indicating the rate at which each terminal node accepts data, which is typically much slower than the rate at which the data is communicated through the network. Node 174 uses this table of information to execute a line allocation algorithm in which a parameter is maintained for each active terminal in the subnet below node 174 which indicates the amount of data stored in buffer storage for each terminal. Each time a data block is sent to a given terminal, the stored data quantity parameter is incremented, and at predetermined time intervals, the stored data quantity parameter is decremented to reflect the rate at which the terminal accepts data from its associated buffer storage. As space becomes available in one of the two private queues in OE96NE 192 and OE96NW 184 corresponding to the two downward paths, microprocessor node 174 searches each buffer pair within large two port 172 to determine if data is available for transmission to the corresponding terminal. If data is available, the parameter which indicates the amount of data stored at the corresponding terminal is retrieved from stroage and compared with other similar parameters. The next data block is thus selected for transmission through the network to the terminal among those for which data is available in large two port 170 which has the smallest backlog of stored data at the terminal as determined by the predictive terminal storage parameters. By maintaining a uniform backlog of data blocks in buffers at the terminal site for terminal nodes for which data is available, optimum use is made of the subnet communication link and short term peak loads which might otherwise cause a response delay at the terminal nodes are accommodated. During times of peak demand, the stored backlog of data at the terminal nodes, coupled with continued transmission at a maximum rate through the network, assures the continuous availability of data to each active terminal unless a period of peak demand has an unusually long duration. This buffering provided by the network storage capacity allows more terminals to use the same communication links without peak demand information delays. The cost effectiveness of the expensive transmission links is thus improved.

Node 174 functions to insert a sequence number in the five least significant bits of the third byte of each block to indicate the cell position of each block within a 16 cell buffer. This assures reassembly of the blocks in the proper order upon receipt at the terminals.

Still another function of line master node 174 is to test and maintain the communication integrity of the 9.6 KB data link. It periodically sends a continuity message, LPCNTM, through data lines 196 and 198. If a continuity message for a given line fails to come full circle back to node 174 three times in a row, a line failure is presumed. Node 174 then begins isolating the failure by commanding a loop back at the farthest line exchange unit device from each end. For example, a message is sent through link 196 commanding a loop back at line exchange unit device 64. Continuity is then tested with a continuity message. If no continuity is found, the next farthest line exchange unit device is commanded to loop back. The farthest device for which continuity exists is thus determined and a loop is established through link 196 to this farthest terminal and back on link 198. Similarly, a longest possible loop is established from the other end. The communication failure is thus isolated and the network continues to communicate with as many terminals as possible until the failure is corrected or repaired.

Micronode LM96N 174 processes messages OTDATM, ALCALM, REINM, NCRDM, NCWRM, NCRSPM, ALCLRM, INDATM, SETAKM, ALMSRM, ACKPRM, LPCNTM, LPBAKM, CLPBKM, REQSTM, and ABORTM.

Large two port 172 includes a large two port adapter module designated LM2PORT, and four 8KRAM modules designated 8KRAMD-3.

Master up queue 228 contains sixteen cells, each storing a 24 byte large block of data. Master down queue 227 stores and passes all downward flowing messages except OTDATM. These include ALCALM, ABORTM, NCRDM, and NCWRM.

Master up queue 228 also provides storage for sixteen cells for large data blocks of 24 bytes each. It stores and passes all upward flowing messages from micronode LM96N 174. These upward flowing messages include messages ALCLRM, PRSNDM, INDATM, and NCRSPM.

The master buffer pairs 220, represent two master buffers 222, 224, each providing storage for sixteen cells, each cell storing a large 24 byte data block of information. The master buffer pairs hold host message output blocks of data for corresponding terminals until the data can be sent through the subnet to a terminal node device. For ease of distinguishing, the buffer pairs are designated MBP followed by a first number, a comma, and a second number. The first number designates a particular one of 32 buffer pairs numbered 0-31 and the second number 0 or 1 indicates a particular buffer of a buffer pair. The master buffer pairs pass message OTDATM.

Four two port buffer memories 178, 182, 186 and 190 implement queues for carrying messages between micronode 174 and micronodes 180, 184, 188 and 192 respectively. Each of these two ports is structurally identical and includes a small two port adapter module and a RAMA[0.25K] module. The two port adapter module for two port buffer memory 178 is designated IE2PORT while the RAM array is designated IE2PRAM. The two port adapter module for two port memory 182 is designated OW2PORT while the memory is designated OW2PRAM. The two port adapter module for two port buffer memory 186 is designated IW2PORT while the memory is designated IW2PRAM. Similarly, the two port adapter module or two port memory 190 is designated OE2PORT while the memory is designated OE2PRAM. Two ports 178 and 186 implementing upward flowing queues EUQE 179 and EUQW 187 respectively are substantially identical and contain storage for ten cells of 24 byte long blocks of data. They pass messages INDATM, SETAKM, ALMSRM, ACKPRM, LPCNTM, ALCLRM, and NCRSPM. Two port buffer memories 182 and 190 implementing downward queues EDQW183 and EDQE191, respectively, are also substantially identical and contain storage for ten cells of large 24 byte data blocks. These two ports carry messages OTDATM, REINM, REQSTM, LPBAKM, CLPBKM, ALCALM, NCRDM, ABORTM and NCRWM.

Micronode IE96NE 180 includes an MPUMOD module designated IE96NEMPU, and NODEADR module designated IE96EADR, a RAMA[0.25K] module designated IE96ERAM, a PROM[2.5K] module designated IE96EPROM, and for some purposes it is considered to include the channel input module 206 and input matching module IM96M. Micronode 180 operates to either pass on or discard all data blocks received through the data link 198. It processes messages ALCALM, LPCNTM, INDATM, SETAKM, ALMSRM, ACKPRM, NCRDM, ALCLRM, NCWRM, and NCRSPM. It provides a conversion from the small 12 byte double block form in which data is communicated over the data link to a 24 byte large block form over which the data is communicated through the network above the data link.

Micronode OE96NW 184 includes an MPUMOD module designated OE96NWMPU, an NODEADR module designated OE96WADR, a RAMA[0.25K] module designated OE96WRAM, and for some purposes is considered to include channel output module 208 and output matching module 205. Micronode 184 moves blocks of data from queue EDQW 183 to a private internal queue for output on the digital data service line 196. Micronode 184 converts the data from a large 24 byte block format to a small 12 byte double block format in which it is moved over the data link. Whenever the private queue contains no 12 byte small blocks, empty frames consisting of a one bit followed by 95 zeros are communicated to channel output module 208 for communication over the data links. Micronode 184 provides synchronous frame timing for the 9.6 kilobit digital service link and for line master micronode LM96N 174.

Micronode 184 processes messages ALCALM, LPCNTM, OTDATM, NCPSPM, ALCLRM, REQSTM, LPBAKM, CLPBKM, ABORTM, REINM, NCRDM, and NCWRM.

Micronode IE96NW 188 includes an MPUMOD module designated 196NWMPU, an NODEADR module designated I96WADR, a RAMA[0.25K] module designated I96WRAM, a PROM[2.5K] module designated I96WPROM and is sometimes considered to contain channel input module 212 designated I96WCHIM as well as input matching module 210. Micronode 188 is functionally identical to micronode 180 as described above.

Micronode OE96NE includes an MPUMOD module designated OE96NEMPU, a NODEADR module designated OE96EADR, a RAMA[0.25K] module designated OE96ERAM, a PROM[2.5K] module designated OE96EPROM, and for some purposes is considered to include channel output module 214 designated OE96ECHOM as well as output matching module OM96M 211.

Referring now to FIG. 1, the 9.6 kilobit line exchange unit devices 62-64 are substantially identical except for the assignment of mutually exclusive addresses to micronodes therein. Each line exchange unit device maintains a table of addresses for addressable micronodes connected therebelow (a terminal device such as device 41 appears to a line exchange unit device as an addressable terminal node within a terminal exchange unit device such as device 73) and selectively removes data blocks from the 9.6 kilobit data links for communication to downstream addressable micronodes or passes the information along the 9.6 kilobit data link to the next line exchange unit device or to line master exchange unit device 60 as the case may be.

Referring now to FIG. 5, 9.6 kilobit line exchange unit device (LEU) 64 is exemplary of the line exchange unit devices and will be described in greater detail. 9.6 kilobit line exchange unit device 64 is coupled to a pair of utility supplied data service units 240, 242 which are identical to the data service units 200, 202 shown in FIG. 4. A loop back module 244 is bidirectionally coupled to data service unit 240 while a loop back module 246 is bidirectionally coupled to data service unit 242. Loop back module 244 selectively operates in a normal mode of operation to receive signals DAIN and C1A from an input matching module IM96M 248 and to pass these signals straight through to a channel input/output module CHIOM 250. Simultaneously, loop back module 244 operates in a normal mode to receive signal C1A and signal DAOUT from channel input/output module CHIOM 252 and pass the two signals straight through to a phase converter module PHCM 254 which synchronizes the data signal with a transmit clock signal C3A received from output matching module OM96M 258 and provides the synchronized data signal to the module 258. The input matching module 248 receives a clock signal C1A and a data signal DAIN from the data service unit 240, inverts the two signals, and passes them on to loop back module 244. The output matching module 258 receives signal DAOUT and inverts it to provide a signal DAOUT to data service unit 240. The output matching module 258 also receives a transmit clock, C3A, which is synchronized with the DAOUT data signal, inverts the transmit clock and communicates the inverted clock to the phase control module 254 as signal C3A.

Loop back module 244 is selectively converted between a loop back mode and a normal mode in response to a 1 or a 0 respectively on data bus line DBO at the occurrence of a loop back write command LPBK active low condition from a micronode D96NW 260. In the loop back mode, loop back module 244 receives signals C1A and DAOUT from channel input/output module 252 and communicates the same signals back to channel input/output module 250.

The coupling of data service unit 242 to line exchange unit device 64 is a substantially identical mirror image of the coupling of data service unit 240. An input matching module IM96M 262 receives signals DAIN and C1A from data service unit 242, inverts these signals, and passes them on to loop back module 246. In a normal mode of operation, loop back module 246 passes these signals on to channel input/output module 252. Loop back module 246 also receives signals DAOUT and C1A from channel input/output module 250 and passes them on to phase converter module 264. Phase converter module 264 synchronizes the data signal DAOUT with a transmit clock signal, C3A received output matching module OM96M 265 and provides the synchronized data signal to output matching module 265. Output matching module 265 receives a transmit clock signal, C3A, from the data service unit 242 which is inverted and communicated to phase converter module 264 as signal C3A. Output matching module 265 also receives the data signal DAOUT from phase converter module 264, inverts it, and communicates it to the data service unit 242 as signal DAOUT for transmission over the data link.

Loop back module 246 is selectively controlled to be in a normal or loop back mode in response to micronode D96NE 266. In the loop back mode, loop back module 246 receives data and clock signals from channel input/output module 250 and communicates them directly to channel input/output module 252.

Line exchange unit control device LEUCNTRL 268 provides the overall control for the other standard modules in line exchange unit device 64. It provides a 1 MHz master clock signals, run halt controls for program debugging, a power on detecter-inhibiter, and a master reset control signal.

Micronode D96NW 260 is connected to receive 24 byte large blocks of data from coax in end node, CIEN 272 through a two port buffer memory 274 which implements a unidirectional queue data up queue west DUQW 276. The data blocks received through DUQW 276 are communicated to the data link through channel input/output module 252. In the present implementation upward flowing messages are always compressible into 12 byte small blocks of data even though they pass through the queue 276 in a large block format, and are given priority over downward flowing blocks of data on the data link.

Upon the storage of a block of data in queue 276, micronode D96NW 260 begins searching for a suitable data frame among the data flowing through channel input/output module 252. Upward flowing messages are allowed to pass therethrough, but upon receipt of the first empty frame or downward flowing message, the upward flowing message stored in queue 276 is inserted into the data stream and occupies one 12 byte time frame therein. If the time frame receiving the upward flowing message previously carried a downward flowing small block of data, the small block is temporarily stored by micronode 260 and inserted inthe next time frame that does not carry an upward flowing message. If the next time frame itself carries a second downward flowing small data block, the second small data block is replaced on the communication link by the first small block and stored until space is available. This process continues until one or more empty time frames appear on the data link to permit all of the temporarily stored small blocks of data to be inserted into an empty time frame. Because of the priority assigned to the upward flowing small blocks, a terminal operator is assured a rapid response because the keyboard requests flow rapidly on a priority basis to host data processor 12, which can process a request and rapidly begin sending downward flowing return data to meet the request. As explained previously, the host data processor 12 can supply downward flowing data faster than it can be received from a terminal and the line allocation algorithm operates to assure optimum utilization of the network data carrying capacity to permit each terminal to receive downward flowing data at its maximum data rate. Optimum terminal usage and minimum operator inconvenience and delay is thereby realized by initiating the downward flow of the return message as soon as possible.

Micronode D96NW 260 also maintains a table of addresses for terminal nodes which are downstream therefrom and upon detecting a message passing through channel input module 252 addressed to one of these downstream terminal nodes, diverts the message from 9.6 kilobit data link through a two port 278 which implements a unidirectional drop down queue west 280 to micronode coax output end node COEN 282, which transmits the data block along a coaxial cable 284 to a downstream terminal node.

Micronode D96NW 260 has a pair of I/O write enable outputs LOADBYTL and LOADBYT2 coupled to channel input/output module 252 to write therein respectively the least significant and most significant bytes of a double byte of information for communication over the data link. Channel input/output module 252 accumulates a serially received double byte of information in a serial in and parallel and serial out shift register and generates an interrupt request IRQ when a full 16 bits are available. To read this information, the micronode 260 responds to the interrupt request by generating a read control signal through an I/O select output with lowest order address bit AO commanding reading first the least significant byte and then the most significant byte. The destination address of a block of data is always contained within the first 16 bits to permit micronode 260 to rapidly determine whether the block of data contains an address for a downstream addressable micronode which should be pulled off the 9.6 kilobit line and passed downward or whether it should be allowed to pass along the 9.6 kilobit line. An I/O read select output, READSRS, commands the reading of the shift registers while a similar output, READCNTR, commands the reading of a counter inside channel input/output module 252 which counts from 0 to 5 as the six double bytes of a small block of data pass therethrough. Micronode 260 may also respond to the interrupt request by substituting selected output data for serially received data in the output portion of CHIOM 252.

Micronode D96NW 260 includes an MPUMOD module designated W96DNMPU, a RAMA[1K] module designated WDNRAM, PROM[4K] module designated WDNPROM, a BUSXTEND module designated WDNBUSX, a NODEADR module designated WDNADR, and for some purposes is considered to include the channel input/output module 252 designated WESTCHIO, the phase converter module 254 designated WPHCM, the input matching module 248 and the output matching module 258. Micronode 260 passes messages ALCALM, LPCNTM, ODTADM, REQSTM, LPBAKM, CLPBKM, ALCLRM, INDATM, ACKPRM, SETAKM, ALMSRM, ABORTM, REINM, NCRDM, NCWRM, and NCRSPM.

Drop 96 node east D96NE 266 is the counterpart of micronode 260 for channel input/output module 250. It receives upward flowing 24 byte large blocks of data (containing 12 byte small block messages) from coax input end node 272 through two port buffer memory 286 which implements drop up queue east DUQE 288. Micronode 266 also receives downward flowing messages from channel input/output module 250 and communicates them through two port buffer memory 290 which implements drop down queue east DDQE 292 to coax output end node 282.

Micronode D96NE 266 is functionally and structurally equivalent to node 260 and includes an MPUMOD module designated E96DNMPU, a RAMA[1K] module designated EDNRAM, a PROM[4K] module designated ENPROM, a BUSXTEND module designated ENDBUSX, a NODEADR module designated EDNADR, and for some purposes is considered to include channel input/output module 250 designated EASTCHIO, phase converter module 264 designated EPHCM, output matching module 265 and input matching module 262. It receives messages ALCAM, LPCNTM, OTDATM, REINM, REQSTM, LPBAKM, CLPBKM, NCRDM, NCWRM, ALCLRM, INDATM, ACKPRM, SETAKM, ALMSRM, and ABORTM, and NCRSPM. Micronode 266 is a switching and injecting synchronous node which serves as part of the line exchange unit 64 attaching to the east to west 9.6 kilobit loop. It switches downward flowing blocks addressed to nodes on the coax loop which it has access to via micronode COEN 282 with switched blocks being passed to COEN 282 for transmission as blockettes on the coaxial cable 284. Micronode D96NE 266 also relays other blocks through CHIOM 250 and inserts upward flowing blocks from the coaxial cable 284, which are received via micronode CIEN 272, onto the 9.6 kilobit loop according to priority and time availability. Micronode D96NE 266 includes a table of downstream micronode addresses permitting it to determine the addresses of downstream micronodes. Micronode 266 also operates in response to control messages addressed thereto and received through the network to control loop back module 246 in a normal or loop back configuration in response to messages from line master exchange unit device 60 to maintain maximum communication capability for the subnetwork in the event of a data link communication failure.

Clock multiplier times six module, CLMX6 294, receives the 9.6 KHz clock signal C1A from input matching module 246 and includes a multiplying phase locked loop which generates clock signals C1B and C2B at 57.6 KHz to serve as master clock signals for the coaxial cable loop 284. Signal C2B leads signal C1B by one-fourth 57.6 KHz bit rate period.

A channel output module 300 includes a 16 bit double byte parallel in serial out shift register which receives parallel data from micronode COEN 282 and responds to the 57.6 KHz C1B clock signal to drive the coaxial cable 284 through an output half 302 of coax transceiver 304 as shown in FIG. 5B. At the other end of the coax loop, data passes through an input half 306 of coax transceiver 304 which communicates the data to a channel input module 308 along with clock signal C1B which permits a conversion by channel input module 308 from a special self-clocking signal format on coaxial cable 284 to the standard separate data and clock format used elsewhere within the communication network. Channel input module 308 includes a 16 bit double byte serial in parallel out shift register and generates an interrupt signal IRQ to micronode CIEN 272 to permit the reading of a double byte of data from the shift register when it is full. A read counter may also be read in response to signal READ CNTR by micronode 272 to indicate the location of a double byte of data within a small block of data containing six double bytes which is the format for communication along the coaxial cable 284. As with communication along the 9.6 kilobit link, 24 byte large blocks of data are actually communicated through the coaxial cable 284 as two 12 byte (6 double bytes) small blocks of data. Each data time frame carries one small block of data with 96 bits.

Micronode CIEN 272 includes an MPUMOD module designated CIENMPU, and NODEADR module designated CIADR, a RAMA[1K] module designated CINRAM, a PROM[4K] module designated CINPROM, and for some purposes is considered to contain channel input module 308 designated CHIN and the input portion 306 of coax transceiver 304 designated 1/2(COAXT/R).

Micronode CIEN 272 is a switching, synchronous 57.6 kilobit micronode which passes on or discards all small blocks of data received from the input end of the coaxial cable 284. It passes all switched small blocks to micronode D96NE266 or D96NW260 in response to a shortest path determining algorithm which uses information derived from the all call messages and all call response messages. The micronodes D96NW260 and D96NE266 insert into the all call messages flowing down therefrom a numerical count parameter which indicates the relative positions of these micronodes between the ends of the 9.6 kilobit data link. This information is used by micronode CIEN 272 to determine the shortest path through micronode 260 or micronode 266 to the line master exchange unit 60 and this shortest path is preferentially selected for the communication of upward flowing data blocks.

Micronode CIEN 272 handles messages ALCALM, ALCLRM, INDATM, ALMSRM, REQSTM, OTDATM, ACKPRM, REINM, NCRDM, SETAKM, ABORTM, NCWRM, and NCRSPM.

Micronode COEN 282 operates to receive data blocks from micronodes D96NW260 and D96NE266 and communicate the received information along the coaxial cable 284 through channel output module 300. Channel output module 300 designates an interrupt request IRQ when a 16 bit latch therein is ready to receive data and micronode COEN 282 responds by generating a signal LOADBYT1 to command the loading of the least significant 8 bit byte into the latches and a signal LOADBYT2(RSTIRQ) to command the loading of a second most significant byte into the latches and to reset the interrupt request.

Micronode COEN 282 is a nonswitching synchronous micronode which moves blocks of data from queues DDQE 292 and DDQW 280 to channel output module 300. The blocks of data are moved into a private queue within micronode 282 for communication on a space available basis and whenever no data blocks are stored within the private queue, empty frames containing a single one start bit followed by 95 zeros are transmitted over the coaxial cable 284. Micronode 282 communicates messages OTDATM, ALCALM, REQSTM, ABORTM, NCRSRM, ALCLRM, REINM, NCRDM and NCWRM. Micronode COEN 282 includes an MPUMOD module designated COENMPU, a node address module NODEADR designated COADR, a RAMA[1K] modules designated CONRAM, a PROM[4K] module designated CONPROM, and for some purposes is considered to include channel output module 300 designated CHOUT and output portion 302 of coax transceiver 304 designated one-half (COAXT/R).

The two port buffer memories 274, 286, 278, and 290 include a S2PTADAPT module designated WO2PORT and a RAMA[0.25K] module designated WO2PRAM. Similarly, two port 274 includes S2PTADAPT module designated WIN2PORT and a RAMA[0.25K] module designated WIN2PRAM, two port 290 includes an S2PTADAPT module designated EO2PORT and a RAMA[0.25K] module designated EO2PRAM while two port 286 includes an S2PTADAPT module designated EIN2PORT and a RAMA[0.25K] module designated EIN2PRAM.

As shown in FIG. 5B, data is transmitted on the coaxial cable 284 in a self clocking mode in which the first quarter of each bit period is always high or logic 1, the second and third quarters reflect the data, and the fourth quarter is always logic zero to guarantee a logic transition to logic one at the beginning of the next bit period. Some of the modules are used for both high speed and low speed lines and in the convention used herein a suffix A refers to a 9.6KB data link and a suffix B refers to a 57.6KB coaxial cable data link.

Input 1/2 coax transceiver 306 provides data signal DBIN and clock signal C1B, which makes a low to high transition at the midpoint timewise of a logic 1 on signal DBIN. Signal C1B may thus be used to clock data input shift registers within CHIm 308.

Multiplier CLMX6 294 derives master clock signals C1B and C2B from signal C1A as provided by loopback module 246. These signals control CHOM 300 and output coax transceiver 302 and provide the master timing on the coaxial cable 284. While the same symbol C1B is used to identify both the clock signal from CLMX6 and the clock signal from input coax transceiver 306 because of their similar relationship to the data signals, it should be appreciated that the two signals will not in general be in phase.

The signals derived from and provided to the 9.6KB line are substantially as shown in FIG. 4B. Clock signal C1A makes a low to high transition at the middle of a DAIN data bit period and the phase converter modules 254, 264 synchronize the phase of DAOUT to place the middle of each data bit period at a low to high transition in signal C3A.

Information passing around the coaxial cable 284 passes through a plurality of terminal exchange unit devices where blocks of data containing addresses for terminal nodes within the terminal exchange unit devices are pulled off and communicated to a terminal device coupled thereto and blocks of data from terminal devices are inserted on the coaxial cable to be communicated back around to the line exchange unit device 64.

Terminal exchange unit device (TEU) 73 is exemplary of the terminal exchange unit devices and, as shown in FIG. 6, includes an input portion 310 of a coax transceiver 312 which receives information from coaxial cable 284 and communicates the information to a channel input/output module 314. Channel input/output module 314 is coupled to a terminal node 316 which selectively receives blocks of data from channel input/output module 314, provides blocks of data to channel input/output module 314 or allows blocks of data to pass through channel input/output module 314 from the input portion 310 of coax transceiver 312 to an output portion 318 of coax transceiver 312 to be communicated along the coaxial cable 284. Terminal node 316 communicates through a two port buffer memory 74, which implements a terminal down buffer queue TDWBQ 320 and an terminal up queue TUADQ 322, with a terminal matching node 324. Terminal matching node 324 is not addressable and is not considered part of the network. For purposes of discussion it is considered part of data terminal 41 and provides a data format interface between the parallel internode data format appearing on the network 10 and the serial or other data format required by most terminals. A terminal interface module 326 provides an interface between the input and output serial data paths of terminal console 328 and the single bidirectional serial data path between terminal matching node 324 and terminal interface module 326.

An exchange unit control module designated TEUCTRL 330 provides the general control functions for terminal exchange unit device 73 including a 1 MHz clock signal, run/halt control for use of program debugging, power up detection, and a master reset signal.

Micronode TN 316 includes an MPUMOD module designated TNMPU, and NODEADR module designated TNADR, and RAMA[1.5K] module designated TNRAM, a PROM[3K] module designated TNPROM and for some purposes is considered to include the channel input/output module 314 designated CHIF and the coax transceiver module 312 designated COAXT/R. Terminal node 316 handles messages OTDATM, ALCALM, INDATM, REINM, REQSTM, NCRDM, NCWRM, ALCLRM, ACKPRM, SETAKM, ALMSRM, ABORTM, and NCRSPM. Node 316 is a switching and injecting synchronous node which interfaces terminal matching node 324 with the coaxial cable 284. It moves 24 byte large blocks of data from terminal matching node 324 onto the coaxial cable in 12 byte small block form and removes 12 byte small block form data from the coaxial cable which is destined for the terminal console 328. From the viewpoint of the network, it serves as the origin and destination for data relating to terminal 41. Terminal node 316 has internal queue storage for two 16 block downward flowing messages to provide a data buffer for information being communicated from host data processor 12 to terminal console 328. Two port buffer memory 74 includes an S2PADAPT module designated TMN2PORT and a RAMA[0.25K] module designated TMN2PRAM.

Two port 74 implements downward flowing TDWBQ 320 which stores three large blocks of data and upward flowing queue TUADQ 322 which stores six large blocks of data as a buffer interface between terminal node 316 and terminal matching node 324. TDWBQ 320 carries message OTDATM and queue TUADQ 322 carries message INDATM.

Only micronode TN 316 carries an address within the terminal exchange unit device 73. Downward flowing data blocks carrying this address are automatically pulled off by terminal node 316 and communicated to the terminal console 328. Similarly, messages from terminal console 328 are configured as data blocks and the address of terminal node 316 is inserted therein before they are communicated upward toward the host data processor 312. Thus, while terminal node 316 is considered to be part of the network and not part of the data terminal, the address of terminal node 316 defines the address of the data terminal connected thereto.

Referring now to FIG. 7, the small two port adapter module 340 is shown as including three 8 bit two to one three state multiplexers 342, 344 and 346 providing communication between two processor 1 and processor 2 data buses and a single random access memory data bus. A single 12 bit two to one three state multiplexer 348 couples the two processor 1 and processor 2 address buses to a single RAM address bus. Multiplexer 348 carries eleven address signals permitting connection of up to 2K bytes of memory and a READ signal for each processor.

Small two port adapter 340 operates to alternately connect a first processor and a second processor to a random access memory during alternate half cycles of a 1 megahertz clock signal having opposite phase components P phase 1 and P phase 2. The select inputs of multiplexers 342 and 348 are coupled to clock signal P phase 1 so that during the half cycle when this clock signal is low, the A inputs which are connected to processor 1 are enabled and during the alternate half cycles, when signal P phase 1 is high, the B inputs connected to processor 2 are enabled and pass through to the Y outputs. Multiplexers 344 and 346 carry data signals from the RAM data bus to the processor 1 data bus and processor 2 data bus respectively. Only the A inputs are coupled to active data signals and they are continuously enabled.

Multiplexer 342 remains in a high impedance output state unless enable input G is driven low by a signal PWRITE in response to the coincidence of a high or write condition on signal RAMREAD (which is generated by a processor that is active during a selected half cycle and passes through multiplexer 348) and a signal RAMENABLE which is active high only when a processor which has been enabled for a given half cycle generates no active signals on any one of the four disable outputs connected thereto. A pair of five input OR gates 350, 352 receive the four disable signals from processor 1 and processor 2 respectively as well as signals P1 and P2 which respond to the clock signals to activate the OR gates during the time when the processor coupled thereto does not have access to the two port memory. Thus, one input to a NAND gate 354 is always enabled and in the event that one of the disable outputs from a processor having access to the two port becomes active, NAND gate 354 becomes fully enabled to prevent generation of signal RAMENABLE which permits reading or writing of the random access memory within the two port.

Only when a processor having access to the two port memory generates a read signal through multiplexer 348 to generate signal RAMREAD does a NAND gate 356 or a NAND gate 358 become fully enabled to generate a signal P1READ or P2READ to enable the output of multiplexer 344 and 346 to permit reading by processor 1 or processor 2 respectively. The communication of information appearing on the RAM data bus to the appropriate processor 8 bit data bus is thus enabled. It is thus apparent that during the positive half cycle of 1 MHz clock signal P phase 1 processor 1 is connected to the random access memory 360 as though processor 2 did not exist. Similarly, during the negative half cycle of clock signal P phase 1, processor 2 is connected to the random access memory 360 as though processor 1 did not exist. Each processor thus has complete access to the random access memory 360 during alternate half cycles of the clock signal P phase 1.

The implementation of RAM array module 360 as presented in FIG. 7 is straightforward with an 8 bit decoder being responsive to address bits 8, 9 and 10 to generate 8 chip enable signals. The RAM array 360 may thus be constructed from modular increments of 256 words with each output of the decoder enabling a different block of 256 words. The small two port adapter module 340 and the RAM array 360 together constitute a small two port buffer memory of the type which is used throughout the communication network 10. The disable inputs are not connected in the small two port configuration but are available for array selectio where multiple arrays are connected to a microprocessor module. The RAM OFF is connectable to the disable signal of the bus extender module.

Referring now to FIG. 8, a standard MPU module, MPUMOD 370 includes an MC6800 microprocessor 372 and a small amount of interface and control circuitry. An in/out write select decoder 374 operates to decode enable and address select inputs to provide an active signal on one of eight address selected outputs when fully enabled. Similarly, an in/out read select decoder 376 responds to the proper combination of address output signals to activate one of eight read select outputs. These write and read select outputs are connected to peripheral units such as channel input modules, channel output modules, or channel input and output modules to command the writing and reading of information relative to these input/output peripheral units.

A memory control decoder 378 responds to the three most significant address bits A13-A15 to divide the addressable memory fields into eight, 8K segments. Of these eight segments, one is connectable to select an internal random access memory array, one is connectable to select a small two port buffer memory to which the microprocessor may be connected, for example, by connection as one of the four disable inputs. Four outputs are available for selection of memory space in a large two port memory such as memory 172, one signal is available to select the input/output decoders 374, 376 and one can select a programmable read only memory. A combination microprocessor reset and deadman's stick is implemented with a pair of timing circuits including an 8 millisecond timer 380 and a 10 microsecond timer 382. An external reset signal sets timer 380 and timer 382 to cause timer 382 to generate a 10 microsecond reset signal for the MPU 372. If within 8 milliseconds, the MPU does not respond to a periodically executed program instruction (for example in case an error has occurred in program execution) by generating an input/output write select 7 signal by executing a dummy write into the corresponding address location, to reactivate timer 380, timer 380 times out and causes timer 382 to generate a 10 microsecond reset pulse. A timer 384 generates a short write gate pulse, WRITEG1 and a write pulse, WRPULSE when signal CLOCK 2 is high while a write operation is commanded by microprocessor 372.

Referring now to FIG. 9, standard module NODEADR 390 provides the mutually exclusive address for a given micronode. The address is defined in one's complement form by a pair of switch sets 392, 394. Switch set 392 contains eight switches which are coupled between ground and each of eight address lines to define a most significant byte of a 16 bit address. Similarly, switch set 394 includes eight switches, each of which is connected between ground and a different address line to define the eight bits of a least significant address byte. An eight bit two to one multiplexer 396 has eight three state inverting outputs which are connected to the data bus of an associated MPU module. An enable input signal, is connected to be activated by an in/out write select output, I/O WSEL from a connected microprocessor module with signal LSBSEL being driven by a bit zero address line of the microprocessor module to select first the most significant byte of the address and then the least significant byte of the address for reading onto the microprocessor module data lines.

Referring now to FIG. 10 a standard module BUSXTEND 400 includes a bidirectional bus driver 402 for the eight bit data bus, NAND gates 404, 406 connected to disable the bus and prevent the reading of information onto the processor data bus except when the read/write processor output is in a read condition, and when at least one of eight disable inputs is at logic zero. For example, the eight decoded address outputs which divide a microprocessor address field into eight parts might each be connected as a disable input. Thus, unless at least one of these outputs is low, information cannot be transferred from the extender bus to the processor data bus. Similarly, a disable input of bus driver 402 is connected to the R/W signal through NAND gate 406 to prevent data from appearing on the extender data bus from the processor data bus except during a read microprocessor operation. A NAND gate 408 receives two input gating signals and generates an output gate signal in response thereto while a buffer 410 receives address signals A0-A9 to provide a fan out of bus extender address signals XA0-XA9 in response thereto.

Referring now to FIG. 11, the exchange unit control module, EUCONTRL 414 includes a 1 megahertz crystal clock generator 416 generating opposite phased clock signals. PRECLOCK1 and PRECLOCK2 and corresponding opposite phase clock signals CLOCK1 and CLOCK2 which lag the corresponding PRECLOCK signal by about 20 nanoseconds. The PRECLOCK signals are used to drive the two port memories to enable them to be ready for accessing by the microprocessors at the time of transitions in the clock signals CLOCK1 and CLOCK2 which drive the microprocessors. A set reset flip-flop 418 drives a JK run half flip-flop 420 in response to a manual switch to permit the selective running or stopping of the modules at an exchange unit. Flip-flop 420 provides a run command only when enabled by flip-flop 418 and the presence of a logic one debug control A signal. When enabled, flip-flop 420 generates at its Q output a RUN signal which changes from low to high with the rising edge of clock signal CLOCK1. This in turn drives flip-flop latches 422 and 424 which generate signals RUN2 and RUN1 respectively. Run halt signal RUN1 is connected to nodes in which clock input signal DATA2 is synchronous with signal CLOCK2 and signal RUN2 is connected to micronodes in which clock input DATA2 is synchronous with clock signal CLOCK1. Flip-flop latch 426 responds to a power on reset signal to provide a synchronous node reset signal.

The programmable read only memory, PROM[0.5K-4K] is not explicitly shown in the drawings because of its conventional implementation. It is implemented as eight 0.5K segments, each eight bits wide. Address bits A0-A8 are connected to select one word from each segment while address bits A9-A11 are connected to a decoder which responds thereto to select one of the eight segments. The decoder has a PROM enable input which may be connected high and two PROM disable inputs, one of which may be connected low and one of which may be connected to the PROMSELECT signal from output Y7 of memory partitioning decoder 378 of a microprocessor unit.

The net master interrupt module is not separately shown because of its simplicity. It includes a divider circuit which receives the 1 Megahertz system clock signal and generates an output clock signal with a 16.4 millisecond period. A presetable four bit count down counter receives the 16.4 millisecond input as a clock signal and counts down toward zero. Upon counting to zero, a borrow output of the counter generates an interrupt signal, IRQ, which is communicated to the netmaster node microprocessor as an interrupt request signal. The interrupt request signal also constrains the divider counter to remain in a reset condition while active and prevents the generation of further 16.4 millisecond clock signals. This condition exists until the microprocessor responds to the interrupt request by activating a load input to a presetable counter through the microprocessor input/output write select output logic to command the loading of the data appearing on the four least significant lines of the data bus into the presetable counter. This terminates the interrupt request and causes the counter to proceed to count down toward zero in response to the 16.4 millisecond clock signals.

The console interface module 430, which is shown in FIG. 12, includes a programmable counter 432 and an asynchronous communication interface adapter 434 of the MC6850 type. Counter 432 receives from the five least significant inputs of a microprocessor data bus 5 bits for presetting the counter. These 5 bits are loaded into counter 432 in response to a load clock signal creseleck from an in/out write select microprocessor output. The five bit count parameter determines the rate at which an output clock signal is generated. This rate is selected under processor control to match the data exchange rate of a terminal console to which the console interface module 430 is connected. The asynchronous communication interface adapter 434 receives the clock signal and provides serial communication through a twisted pair connected receive data line and transmit data line. An eight bit parallel data input/output is connected to the microprocessor 8 bit data bus and an interrupt request signal is connected to the interrupt request of a connected microprocessor. An enable signal designated E is connected to the phase 2 clock input of the console interface node microprocessor while the read/write input is connected to the read/write output of the microprocessor. Address bit A0 drives the register select input to select one of two read registers and one of two write registers. Chip select inputs CS0 and CS1 are connected to +5V while chip select input CS2 is connected to an I10 read select microprocessor output to control the accessing of the asynchronous communication interface adapter 434.

The terminal interface module is shown in FIG. 13 to which reference is now made. Terminal interface module 446 provides bidirectional multiplexing between a line of the data bus such as line DB0 of a terminal matching node microprocessor and a separate input and output for a terminal. Data appearing on the data bus line may be written into a latch 448 in response to an active low hold data signal which may be conveniently connected to a write in/out select output of a connected terminal interface microprocessor. Latch 448 then holds the data and drives the terminal with the appropriate data signal. A read disable signal RDISABLE may similarly be connected to a discrete output of an in/out read select decoder of a terminal interface microprocessor to permit a data signal generated by a connected terminal to appear on the selected line of the data bus. Input and output serial communication with a connected data terminal is thus provided through the terminal interface module 446.

FIG. 14 illustrates the 57.6 KB coax transceiver 454 with FIG. 14A illustrating a receiver portion 456 and FIG. 14B illustrating a driver portion 458. Signals which are useful in understanding the operation of the transceiver 454 are illustrated in FIG. 5B.

Referring now to the input portion 456 shown in FIG. 14A, a receiver having a voltage level biasing circuit 460 connected thereto is coupled to the input of a phase locked loop 462 which is conventionally connected to provide an output signal at twice the 57.6 kilobit data frequency on the coaxial cable. This output is connected to clock a toggling flip-flop 464 at negative transitions thereof and a flip-flop 466 on positive transitions. The Q output of flip-flop 464 drives the D input to flip-flop 466 and is complemented to provide clock signal C2B. The Q output of flip-flop 466 provides a signal C2B.

These clock signals are illustrated in FIG. 5B and serve to divide each data bit period into four intervals. Data is presented on the coaxial cables in a self clocking format in which the first one-fourth of a data period is always at logic one, the second and third fourths of a data period reflect the data content of the bit being transmitted, and the fourth quarter of a data period is always at logic zero. A low to high transition always occurs therefor at the beginning of each data period. Signal C2B goes high during the second and third quarters and is ANDed with the received data signal to generate signal DBIN which indicates the received data during quarters 2 and 3 and is otherwise at logic zero.

Referring now to FIG. 14B, the output portion 458 of the coax tranceiver 454 receives data to be transmitted as signal DBOUT. An AND gate 468 receives signal C2B and the data signal which is phase shifted by one-half bit time period relative to the data input. The output therefrom is connected to an OR gate 470 to drive the OR gate with the appropriate data signal during bit time quarters two and three. An AND gate receives a signal C1B which is ANDed with signal C2B, to provide as an output from AND gate 472 a logic one during the first one-fourth of each bit time interval. This output is coupled as an input to OR gate 470 so that the coax output signal is always at logic one during the first quarter of a bit time interval, reflects the data provided through AND gate 468 during quarters two and three, and is at logic zero during quarter four. This conforms to the required data format.

The clock multiply module, CLMX6, is not explicitly shown but is substantially identical to the portion of FIG. 14A which includes the phase locked loop 462, flip-flop 464, and flip-flop 466. One change is that the input of flip-flop 462 is connected to the low frequency 9.6 KHz signal C1A. A divide by six counter is coupled between the Q output of flip-flop 464 and the compare input of phase locked loop 462 to reflect the frequency multiplication between the input and output of the phase locked loop 462.

Because of propagation delays along a transmission link, data received over a first line may not be in phase with data received over a second line. Nonetheless, it must be provided to the second line synchronously with the clock signal for the second line. The phase converter module 472 shown in FIG. 15 provides the synchronization of the data to be transmitted with the transmission clock signal, C3A for that data line. The phase converter module 472 shown in FIG. 15 includes a D-type edge triggered flip-flop 474 which receives the data input from a channel output module or channel I/O module and provides a data output synchronously with the digital data service transmission clock signal C3A.

Flip-flop 474 is clocked by the transmission clock signal C3A. It thus enters data at the beginning of each transmission bit time interval and holds that data until the beginning of the next transmission bit time interval. A flip-flop 476 has the D input thereof connected to signal DAOUT and is clocked by signal C1A at the middle of each output data valid interval to insure that the data supplied to the phase converter module 472 is held until it can be loaded into flip-flop 474. A series of exclusive OR gates are responsive to signal C1A and signal C3A to load flip-flop 474 either directly from signal DAOUT or indirectly from flip-flop 476. A delay is provided along the series connected Exclusive-OR gates to insure that the loading path for flip-flop 474 is determined according to the relative phase relationships of the two clock signals shortly before the low to high transition in signal C3A. If shortly before the low to high transition of signal C3A, signal C1A is at a high state, flip-flop 474 is loaded with the contents of flip-flop 476. On the other hand, if signal C1A is at a low state shortly before the occurrence of the low to high transition signal C3A, flip-flop 474 is loaded directly through its D input with the data signal appearing on DAOUT. Phase converter module 472 thus provides as an output a a signal DAOUT which is synchronized with the transmit clock C3A notwithstanding phase differences between signal C1A and signal C3A.

The loop back module is not separately illustrated because of its simplicity. It includes a four channel two to one multiplexer and a D-type flip-flop. The D-type flip-flop has the data input connected to a microprocessor data bus signal such as DB0 and its clock input connected to a microprocessor in/out write select signal. The outputs of the flip-flop drive the select inputs of the multiplexer to select alternately channel A when the flip-flop is reset and channel B when the flip-flop is set to place the loopback module in a loopback mode.

The multiplexer has input A1 connected to the data signal DAOUT from a first channel I/O module, input A2 connected to clock signal C1A from the first channel I/O module, input A3 connected to data DAIN from the data link, which in general will not be in phase with the C1A clock signal from the channel I/O module. Outputs D1 and D2 are coupled to the data link while outputs D3 and D4 are coupled to a second channel I/O module. Thus, during a normal, non-loopback mode of operation data and clock signals pass from the data link. In addition, inputs B1 and B2 are connected to signals DAOUT and C1A from the second channel I/O module while inputs B3 and B4 are connected to logic 0. Consequently, during a loopback mode of operation the data and clock signals from the second channel I/O module are returned as inputs to the first channel I/O module. For loopback module 244 in FIG. 5A, CHIOM 254 is the first channel I/O module and CHIOM 252 is the second channel I/O module. The first and second CHIOM's would be reversed for loopback module 246 with module 252 being the first and module 254 being the second.

An understanding of the operation of the channel input/output module (CHIOM) 490 is best had in conjunction with FIG. 16A and FIG. 16B to which reference is now made. The channel input/output module 490 has a synchronization circuit 492, a data input circuit 494, a data output circuit 496 and a timing and control circuit 498. The synchronization circuit includes a sync bit counter 500 and an error counter 502 while the timing and control circuit includes a control bit counter 504 and a D byte counter 506. The synchronization circuit operates to maintain and automatically synchronize the control bit counter 504 and D byte counter 506 with the incoming data stream. As shown in FIG. 16B, the incoming data stream appears as signal DAIN or DBIN depending on whether the module 490 is connected to the 9.6KB data link or the 57.6KB coaxial cable data link. The incoming data formats are nearly identical except for the frequency and for the fact that 1's appearing on signal DAIN occupy substantially an entire bit time interval while 1's appearing on signal DBIN occupy only quarters 2 and 3 of each bit time interval. However, master clocking occurs on the rising edges 508, 509 and 510 of the clock signals C1A and C1B which appear at the center of each bit time interval so that the presence or absence of a logic 1 signal during quarters 1 and 4 of a bit time interval is immaterial.

Synchronization is effected in response to empty frames carrying a logic 1 start bit followed by 95 0's and is not affected by any other data frame. Upon receipt of a logic 1 start bit sync bit counter 500 is preset to 95 through its synchronous parallel load enable input at the rising edge of a clock pulse such as rising edge 510. Any logic 1's appearing in data stream cause sync bit counter 500 to again be preset to 95 and remain ineffective. However, if an empty frame appears with 95 consecutive 0's, sync bit counter 500 counts down to reach a count of 0 upon the occurrence of a clock pulse transition 509. As a result, 0 detect output ZD goes active low during the last half of bit time 95 and during the first half of bit time 0 for the next start bit when counter 500 is clocked by rising edge 510 and preset to count 95. Signal ZD and the inverted data input signal are applied to a NOR gate 512 so that during the first half of the start bit following an empty frame the output of NOR gate 512, which is designated error counter enable 1 (ECEN1), goes high to enable a CET input to error counter 502 at the occurrence of a rising edge 510 of the clock signal.

If the control bit counter 504 and D byte counter 506 are properly synchronized, they reach a count of 15 and 5 respectively upon the occurrence of a clock transition 509 in the middle of bit position 95 to cause a NOR gate 514 to generate a logic 1 end frame signal. The concurrence of the end of frame signal and signal ECEN1 fully enable a NAND gate 516 causing it to generate at its output a logic 0 error counter enable 2 signal (ECEN2) which is coupled to activate the low active parallel enable input to counter 502 and cause counter 502 to be present to a count of 12 at the occurrence of clock transition 510 in the middle of the start bit for the next data frame. Timing and control circuit 498 then maintains control of the channel input/output module 490 operation until the next empty frame permits a test of synchronization.

In the event that the timing and control circuit 498 is not properly synchronized at the end of an empty frame, during the last half of bit 95 and the first half of the start bit for the next data frame, signal end frame will be at logic 0 instead of logic 1 and NAND gate 516 will be partially disabled to generate a logic 1 signal ECEN2. As a result, at the occurrence of the transition 510 in the clock signal, both the CEP and CET enable inputs to counter 502 are enabled and the parallel load enable input is disabled. Clock transition 510 thus causes error counter 502, which was previously set to count 12, to increment to count 13. Upon the occurrence of subsequent empty data frames while the timing and control circuit 498 remains unsynchronized, error counter 502 continues to increment to count 14 and then to count 15. Terminal count output, TC, of error counter 502 is enabled by input CET and is thus immediately disabled after error counter 502 is incremented to count 15 as signal ECEN1 goes low at the middle of the start bit for the next data frame. There is thus no effect on the timing and control circuitry 498 until the occurrence of a fourth empty frame. After the occurrence of a fourth empty frame signal ECEN1 goes high during the first half of the following start bit to enable the TC output of counter 502 so that at the occurrence of transition 510 following a fourth empty frame, signal EC sync is high to enable the synchronous reset input to control bit counter 504 and the parallel enable input to D byte counter 506, which is connected to be preset at 0. Simultaneously, the preset enable input to error counter 502 is enabled by signal EC sync for the fourth data frame and the error counter is preset to a count of 12. The timing and control circuit 498 is thus synchronized and normal operation may continue.

The synchronization circuit 492 thus permits automatic synchronization upon connection of the module 490 into a data stream or upon the loss of synchronization due to a temporary error condition. There is no need for manual intervention. At the same time, this automatic synchronization is accommodated with no need for special synchronization signals which increase hardware requirements or decrease the maximum data flow rate. Instead, the empty frames which inevitably occur in a statistically loaded network which must be capable of accommodating peak loads greater than average are used to effect automatic synchronization.

The operation of the data input circuit 494 is quite straightforward. This circuit includes a 16 bit serial in-parallel out shift register 518 which includes a latch for each shift register bit position, the latches being loaded in response to a strobe signal input from the Q output of a strobe flip-flop 520. The input circuit 494 also includes a multiplexer 522 coupled to output on the 8 bit data bus first, the most significant byte of the latched output of shift register 518 when address bit A0 is at logic 1, and then the least significant byte from the latched output of shift register 518 when the associated microprocessor address bus is incremented to force address bit A0 to logic 0. As the data inputs appear on signal DAIN, they are sequentially loaded into shift register 516 as a positive transition occurs in clock signal C1A at the middle of each incoming bit time interval. The contents of the shift register 518 thus trail the incoming data stream bit time intervals by one-half period and are synchronous with the bit time intervals indicated by control bit counter 504. As each 16 bit double or D byte of data is loaded into shift register 518, control bit counter 504 reaches a count of 15 and generates a TC output signal which clocks flip-flop 520. Flip-flop 520 is connected to always become set when clocked and generate a Q output to drive the strobe input of shift register 518 and load the contents of the shift register into the output latches. The inverter clock signal is connected to the reset input of flip-flop 520 so that at the subsequent high to low transition of the clock signal one-half bit time later, flip-flop 520 is reset to remove the strobe signal and cause the output latches of shift register 518 to retain the originally strobed 16 bits as an additional 16 bit double byte of data is loaded serially into shift register 518. The output of shift register 518 thus represents the last complete double byte of data received on signal DAIN.

The output data circuit 496 includes a 16 bit parallel in-serial out shift register 524 and a pair of latches 525, 526. The serial or parallel loading of shift register 524 is controlled by a pair of flip-flops 528, 529. If during the course of processing a double byte of data, data is loaded into latches 525 and 526 for output, flip-flop 528 is set simultaneously with the loading of latch 525. Subsequently, upon the control bit counter 504 reaching a count of 15, a signal bit count 15 (BC15) is generated shortly after the rising edge of a clock pulse such as rising edge 509. This signal BC15 clocks both flip-flop 528 and 529 thereby resetting flip-flop 528 and transferring the prior contents thereof to flip-flop 529. If flip-flop 528 had not been previously set by an I/O write operation, a 0 is loaded into flip-flop 529 and the Q output drives one input to a NOR gate 530 whose resulting logic low output drives a parallel enable input to shift register 496 and causes the shift register to load serial data from the output of input shift register 518. In this mode of operation, the data input is clocked straight through shift register 518 and then shift register 524 to drive the data output signal, DAOUT or DBOUT. However, if, as a double byte of data is being shifted through the shift registers 518 and 524, data is loaded into the latches 525 and 526, signal BC15 clocks a logic 1 into flip-flop 529 to drive the Q output to 0. Simultaneously, the normally high signal BC15 goes to logic 0 during bit count 15 to fully disable NOR gate 530 and cause its output to go high during bit count 15. Thus, when a clock transition such as transition 510 occurs at the end of control bit counter interval 15 the parallel input to shift register 524 is enabled and data is loaded in parallel into shift register 524 from latches 525 and 526. As control bit counter 504 overflows from count 15 to count 0 upon the occurrence of clock transition 510, signal BC15 again goes high to drive the output of NOR gate 530 low and command serial operation of output shift register 524.

An interrupt request flip-flop 532 generates an interrupt request signal, IRQ, at the Q output therefrom. Flip-flop 532 has its D input connected to logic 1 and its clock signal connected to the terminal count output, TC, from control bit counter 504. Thus, as control bit counter 504 overflows from count 15 to count 0, the signal TC makes a low to high transition to load a logic 1 into flip-flop 532 and indicate to an associated microprocessor that a new double byte count interval has begun. The associated microprocessor then has 16 bit times to reset flip-flop 532 by an I/O read operation and to load a double byte of data into latches 525 and 527 if it is to do so.

D byte counter 506 is coupled to be incremented as control bit counter 504 overflows from count 15 to count 0. D byte counter counts from 0 to 5 and then overflows back to 0 again. Since each data frame interval contains 12 bytes of data or 6 double bytes, D byte counter 506 maintains a double byte synchronization of the bit frames.

The contents of the D byte counter as well as the contents of the input data shift register 518 are available to be read by the I/O controls of an associated microprocessor. For example, when a drop node D96NE or D96NW is connected to the channel input/output module 490, the microprocessor contained therein can read the contents of the D byte counter 506 by executing a read operation at address location C004 (HEX). This generates a signal read counter, READCNTR which causes Q outputs Q0, Q1 and Q2 of D byte counter 506 to be placed on data bus lines DB0, DB1 and DB2 respectively.

Similarly, by reading address locations C000 and C001 the most significant and least significant bytes respectively are read from input shift register 518 through multiplexer 522. As these two address locations are read the input/output circuitry generates a signal read shift registers READSRS which enables the multiplexer 522 to drive the data bus with 8 bits of data. As address location C000 is read address bit A0 is low to enable the A input to multiplexer 522 and cause the most significant byte of data (the first byte of a double byte to be received) to be output through multiplexer 522 onto the data bus. Subsequently, as the address is incremented to C001, output A0 goes high to enable the least significant byte to be output onto the data bus.

The associated microprocessor thus has the capability of reading the D byte counter 506 to determine which double byte of information is contained in the latched output of shift register 518 and then reading both bytes of data from the latched outputs of shift register 518. It should be kept in mind that as a first double byte of data is loaded into shift register 518 and then transferred into the output latches, D byte counter 506 is incremented from 0 to 1. Thus, the count of 1 in D byte counter 506 means that the first double byte of data is being read from the shift registers while a count of 0 indicates that a last double byte of a small block of data is being read from shift register 518.

The input/output circuitry of the associated microprocessor also controls the writing of information into the latches 525 and 526 for subsequent loading into output shift register 524. The writing of data into address location C011 generates an I/O write output which is connected to generate signal load byte 1, LOADBYT1, which loads the contents of the data bus into latch 526. A subsequent write by the microprocessor into address location C012 generates a load byte 2 signal, LOADBYT2, which transfers the contents of the data bus into latch 525 and simultaneously sets flip-flop 528 to command the parallal loading of the data into shift register 524 at the beginning of the next double byte interval.

Referring now to FIG. 17, the channel output module (CHOM) 540 includes a 16 byte parallel in-serial out shift register 542, a pair of 8 bit latches 544, 546, a 4 bit counter 548 and a pair of interrupt control flip-flops 550, 552. Bit counter 548 responds to the master input clock signal to define 16 bit double byte output data intervals. Upon reaching a count of 15, a terminal count output, TC, goes high to present a logic 1 input to flip-flop 550 and enable the parallel load input to shift register 542. At the next occurrence of a low to high transition in the clock signal, bit counter 542 overflows from 15 to 0 to begin a new double byte data interval while flip-flop 550 is clocked to the set state and data is loaded in parallel into shift register 542 from the latches 544 and 546. As flip-flop 550 is set its Q output is connected to the clock input of flip-flop 552 to clock that flip-flop to the set state. The Q output thus goes active low to generate an interrupt request signal, IRQ. As counter 548 overflows to 0, the TC output returns low and the next clock pulse returns flip-flop 550 to the reset state while flip-flop 552 remains unchanged.

A connected microprocessor responds to the interrupt request signal by writing data into latches 544 and 546 using its intput/output circuitry. Selection of an appropriate write output address generates a signal LOADBYT1 which selects most significant byte latch 546 to load the contents of the data bus into the latch. Next, the microprocessor selects another write output address to generate a signal LOADBYT2 to load the least significant byte from the data bus into least significant byte latch 544 and reset flip-flop 552. At the end of the current double byte bit time interval, as indicated by signal end of D byte generated by the TC output of bit counter 548, the previously loaded contents of latches 544 and 546 are transferred to shift register 542 and a new interrupt request is generated.

The input matching module, IM96M, merely provides logic inversion and voltage level buffering between the RS232 DI and CI inputs and the TTL corresponding D0 aND C0 outputs. Similarly, the output matching module, OM96M, merely provides logic inversion and logic level buffering between the RS 232 CI input and the TTL C0 output and between the TTL DI input and the corresponding RS 232 D0 output. Because of the comparative simplicity and conventional implementation of the input matching module and the output matching module these two modules are not explicitly shown in detail.

The channel input module may be identical to the channel input/output module 490 with the data output portion 496 thereof deleted. In fact, except for the extra expense of unused circuitry, the channel input module such as module 206 or module 212 in FIG. 4A could be implemented as a channel input/output module 490 with the output portion simply not connected. Because of the similarity between the channel input modules and the channel input/output module 490, the detailed implementation of the channel input module has not been separately shown and described.

A maximum size 8K by 8 RAM array module 560 is illustrated in FIG. 18. RAM array 560 includes an 8 by 8 array of 1K by 1 random access memory circuits 562, an 8 bit decoder 564 and an 11 bit inverter 566. The array 562 receives 8 bit parallel data inputs DI0-7 and provides 8 bit parallel data outputs D00-7. It receives inverted address inputs A0-9 through inverter 566 as well as a read write signal R/W from inverter 566. The chip enable inputs for each set of 8 parallel RAM 1K memories is connected to one of the Y0-Y7 outputs from decoder 564. The three inputs to decoder 564 are driven by address signals A10-A12 while select inputs G2A and G2B are driven by address signals A13 and A14. A select input G1 is driven by a RAM enable signal which may be driven by a RAM enable output signal from a large 2 port adapter module or may be coupled to logic 1 or to a memory selection output of a microprocessor module. Word sizes smaller than 8K may be implemented by simply omitting a portion of the memory circuits in the array 562.

Referring now to FIG. 19, there is shown therein a large 2 port adapter module (L2PADAPT) 576. This module includes 3 two to one 8 bit multiplexers 578, 580, 582 and a 12 bit two to one multiplexer 584. Because of fan out limitations, the attached 32K.times.8 bits of random access memory are divided into 2 16K by 8 sections, each of which has data outputs driving the A inputs and B inputs respectively to multiplexers 578 and 582 for processor 1 and processor 2 respectively. The select inputs of these multiplexers determine whether the first or second 16K portion of the RAM is selected for reading and the enabled inputs of these multiplexers determine whether or not the information read from the RAM is placed on the processor 1 data bus or the processor 2 data bus. The RAM portions are selected in response to address signal A14 while multiplexer 582 is selected to drive processor 2 in response to a signal P2 driven that is generated in response to the concurrence of a write command and the active low enabling of processor 2 large 2 port select output signals while the phase 2 clock signal is high. Similarly, multiplexer 578 is selected to drive the processor 1 data bus in response to signal P1 driven upon the concurrence of a write pulse, a processor 1 active low large 2 port select output and a high phase 2 clock signal condition. During a write operation, a port select signal determines whether the memory input data bus and address bus is driven from the P1 processor or the P2 processor.

The processor input interface module 100 and processor output interface module 102 have not been explicitly shown in detail. These two modules merely provide an interface between the data format of the communication network and the data format of the host adapter for the host processor. The internal construction of these two modules does not represent any patentable novelty and is readily implemented by a person of ordinary skill in the art. The construction of these modules will depend upon the data format of the host data processing system, but in the present implementation they provide a conversion between an 8 bit per byte 24 byte block format used in the network and a 12 bit byte parallel format used by the host adapter.

All messages, whether for internal control or for data communications between a terminal and a host, use the same standard data format. The data is arranged in the form of blocks containing 24 8 bit bytes. Of these 24 bytes 4 are utilized for overhead and distribution control while 20 contain data that is being transmitted. A message may contain from one to sixteen blocks. A maximum message length is thus 320 bytes of data. If more than 320 bytes of data are to be sent through the network for a given task, it is merely sent as a plurality of sequential messages.

Each block of data contains 24 bytes designated 0-23. Each byte contains 8 bits designated 7-0 going from most significant on the left to least significant on the right. Upon serialization, the lowest numbered byte of a block is sent first and within a byte the most significant or highest numbered bit is sent first. Thus, upon serialization of a block of data for transmission through a communication link, byte 0, bit 7 is sent first followed by byte 0, bit 6 with byte 23, bit 0 being sent last. Byte 0 is considered the least significant and byte 24 the most significant.

In each block of data, the first three bytes occupying positions 0-2 carry header or control information which will be described in greater detail below, the next 20 bytes occupying positions 3-22 carry data information and the final byte occupying position 23 carries an error detecting check sum. This check sum represents a simple byte by byte edition of the information contained within bytes 0-22 with all of the carries or overflows from the most significant bit 7 position being simply discarded.

This 24 byte format provides relatively high efficiency data communication with data being communicated through the network occupying 83% of the total information communicated through the network. However, to further improve communication efficiency in situations where less than a full 20 bytes of data are to be contained within a given block and to permit faster communication of data from a terminal to the host to improve the speed of response of the host, 24 byte blocks of data are broken down into 12 byte small blocks of data for serial communication through the serial data links. If a large 24 byte block carries more than 8 data bytes it is transformed into 2 small blocks of 12 bytes each and if it carried 8 bytes of data or less it is transformed into a single 12 byte small block of data. This use of double small blocks has little effect on large messages where most of the data positions are filled, but doubles the communication efficiency of small messages containing 8 bytes or less because only 12 bytes rather than 24 bytes need be sent through the serial communication links for a given message. At the same time, the basic information handling circuitry and software can be utilized to handle single and double small blocks of data at very little additional cost. The use of small blocks of data on the serial communication links is particularly advantageous for applications such as the keyboard entry of information wherein each message may carry only a single or double byte of data including a keystroke entered by an operator. First of all, only 12 bytes of data are required to carry the keystroke instead of 24 and second, the upward flowing small block may be inserted on the communication link between first and second small blocks of a large block of data. The maximum delay before starting a keystroke containing block of data upward from a terminal toward the host is a 12 byte communication period rather than a 24 byte communication period. A shorter time delay is thus encountered before the host is able to receive the keystroke data and respond with the beginning of a downward flowing message.

When a 24 byte large block of data is transformed to a double small block format for communication along the serial data links, the third byte of control or header information occupying byte position 2 is moved to byte position 12 and all of the bytes in byte positions 3-12 are moved ahead by one byte position. That is, byte 3 is moved up to position 2 which was formerly occupied by the third control byte, byte 4 is moved to position 3, byte 5 is moved to position 4 and so forth up to byte 12 being moved to position 11. Bytes 13-23 remain unchanged except that bytes 12-23 now represent a separate, second small block of data and are renumbered as bytes 0-11 respectively in the second small block.

If a 24 byte large block of data is to be transformed to a single 12 byte small block of data, this transformation is accomplished by merely transforming the check sum in byte position 23 to byte position 11 and then discarding byte positions 12-23. Since the empty data bytes that are discarded carry all zeros, no change in the valid check sum results from moving it from position 23 to position 11 at the end of the single small block of data.

The first header byte, which occupies byte position 0 in a large block or a first small block of data, always contains a start bit of logic 1 in bit position 7. It serves as a start bit for synchronization and detection circuitry which receives the information from along the serial data communication links. The next bit, at byte position 0, bit position 6, is designated FSTBSF or first small block signal field. It is always 0 since this byte is always contained within the first small block of data of a data block even if a second small block of data is present. Bit 5 is the block type indicator signal field designated BTISF with a logic 0 indicating a single small block data format and a logic 1 indicating a double small block format. Bit 4 is the up-down flag signal field designated ABFSF with a logic 0 indicating a type A upward flowing message and a logic 1 indicating a type B message flowing downward from a host toward a terminal. Bit position 3 is the supervision data flag signal field designated SDFSF with logic 0 indicating a data carrying message and logic 1 indicating a supervision type of message for intrnal use by the network. A supervision type of message carries an extra header byte in byte position 3. This extra byte carries a supervision type number which identifies one of many different types of supervision messages. Independent numbering systems are used for input and output messages. Bit position 3 is an unassigned spare designated SPRSF which is available for future expansion of the network, such as by increasing the address field. Bit positions 1 and 0 carry the most significant bits of a node address field designated NASF. Each addressable node is assigned a 10 bit address to permit up to 1022 (address 0 and 1023 are not used) nodes to be individually addressed within the network. Any time a node initiates a message, it includes its own node address within this address field. Thus, in upward flowing messages the address field indicates the source of the message and in downward flowing messages the address field identifies the destination node for the message. A data terminal is not assigned an address as such, but is uniquely identified by the address of a terminal node connected thereto. Thus, within the network, messages to or from a terminal are actually communicated to or from the associated terminal node with the terminal node being coupled to communicate the information to the terminal.

The second header byte in byte position 1 carries the 8 least significant node address bits 7-0 in bit positions 7-0 respectively of byte position 1. The third header byte occupies byte position 2 in the large block format and single small block format but occupies the first position designated byte position 0 of the second small block of a double block format. The most significant bit 7 is the last small block start bit signal designated LBSTSF and is logic 1 to insure a logic 1 first bit for each second small block of data. Bit position 6 is the second small block signal field designated SECBSF and is always at logic 1 to identify the second small block of a small block pair and a double small block format. These first two bits of byte 3 are superfluous when used in the third byte of a single small block format block of data. Bit 5 of the third header byte is the end of message signal field designated EOMSF. Logic 0 indicates that following data blocks are a part of the same message while a logic 1 indicates that the present large block, whether transformable as a single or double small block is the last block of a given message. Bit positions 4-0 contain a block sequence number designated BSNSF. The block of a given message are numbered sequentially from 0 up tp 15 with the sequence number being inserted in these bit positions 4-0. This sequence number permits the several blocks of a message to be assembled in the proper order by a destination terminal node even if the blocks get out of sequence as they pass through the network. The maximum 16 block message size of the present network implementation requires only a 4 bit sequence number with the fifth or most significant bit in bit position 4 of the third header byte being available for future expansion.

The 3 byte header structure permits the selection of a large number of different message types by changing the states of the variable bits within the header. A brief description of the various messages and their header bit states follows.

There are four types of messages which reach the host data interface between network controller 50 and host adapter 32 as shown in FIG. 1. An abort message designated ABORTM carries messag No. 32 in the output supervision type field at byte positions 3 and carries no parameter in byte positions 4-22. The abort message has data states 1, 0, 0, 1, 1, 0 in byte 0 bit positions 7-2 respectively, and bit states 1, 1, 1, 0, 0, 0, 0, 0 in byte position 2, bits 7-0, respectively. The node address is the address of a terminal node for which a message stream is being aborted.

The abort message is issued by the host data processor 12 to abort messages in the process of being output through the network 10. The abort message is related through nodes POMN 100, YN 110, fan-out 54, 456 and B96BN 170 and then through master down queue 228 to node LM96M 174. Mode LM96M 174 responds by initializing the data stream records pertaining to the message being aborted, including the discarding of data in the associated buffer pair 220 by sending the host 12 a permission to send message, PRSNDM, and by relaying the abort message to the destination terminal node. Node LM96M 174 also sets a time-out timer and, until the time-out expires, discards all upward flowing traffic from the destination terminal node. The time-out period is 0.15 seconds which is sufficiently long to permit the discard of all supervision messages that are generated by the destination terminal node before receipt of the abort message.

Upon receipt of an abort message, a destination terminal node sets a 0.20 second delay timer. During the delay period any out of order data blocks reach the terminal node and at the end of the delay period any data blocks in the terminal node buffer pair are discarded and a normal acknowledge and permission to send message, ACKRPM, is sent to the node LM96N 174. Upon receipt of the acknowledge and permission to send message by node LM06N 174, tranmission of a new block set (if any) in a dedicated master buffer pair 220 may proceed. The nodes intermediate the host interface and node LM96N 174 and intermediate node LM96N 174 and the destination terminal node merely pass on the abort message in the normal manner.

The input data message is a data message which flows upward from a terminal node to the host interface. The input data message contains in byte position 0 bits 7-2 bits 1, 0, x, 0, 0, 0,. The x is in the block type identification signal field and indicates whether the input data block should be serially transmitted as a single or double short block and in the present implementation is always 0 to indicate a single short block. The node address contains the address of the originating terminal node and the third byte contains 1, 1, 1, 0, 0, 0, 0, 0 in bit positions 7-0 respectively. Since only single block messages are sent upward through the network in the present implementation, the end of message signal field is always set to 1 and the block sequence number is always 0. Up to 8 bytes of data may be carried in byte positions 3-10 with the check sum being added in byte position 11 for the single small block format and in byte position 23 for the large block format.

The input data message originates in the terminal matching node of a terminal exchange unit and contains up to 8 bytes of data from the terminal, for example, one or more key-strokes. It flows with highest priority through the network from the terminal matching node to the Y node 110. At the Y node it is duplicated and the copy is placed in QHDAUQ 113 for transfer to processor input interface matching node 100 by processor input matching node 104. A second copy if placed in quene NMINQ 137 for net master node 134. The present application is intended for use with keyboard terminals having a relatively low volume data flow upward through the network. The maximum message input data message size is thus limited to 8 bytes, but in general it will be appreciated that for applications where a higher volume of data flow is required, full capacity messages with 16 or more large blocks of data may be utilized. The use of larger upward flowing messages would best be implemented using the same type of message error control and buffer storage that is used for downward flowing messages.

The format of the output data message is substantially the same as the input data message except that up to 20 bytes of data may be contained in each large block and up to 16 large blocks my be contained in each message. Byte 0 bit 5 is set to 0 to 1 to indicate a single or double small block format and bytes 0 bit 4 is set to 1 to indicate a downward flowing message, which in the present system has a lower priority than upward flowing messages. The node address contains the address of the destination terminal node and the end of message signal field at byte position 2 bit 5 carries a 0 or a 1 depending on whether the particular block is the last block of a message. The block sequence number reflects the sequential order of the block within a message. All blocks of a message except the last block are assembled as large blocks containing 20 bytes of data and pass through the serial data links in a double small block format. The format of the last block of a message depends upon the number of data bytes carried thereby as explained previously.

The output data message carries an ordered set of data bits moving from the host 12 to a terminal. The network assures integrity and proper ordering of the bits for each message. The host source must be programmed to pack the data in the proper format for a given destination terminal and it may be necessary for the data bytes to carry information which is interpreted by the terminal as control information indicating such things as the number of bytes in a message, beginning of message, end of message and so forth. The network merely functions to communicate the data portion of a message (exclusive of the header and check sum bytes) to a terminal as prepared by the host. The network is insensitive to the packing of the data bytes within an output data messsge or input data message.

The permission to send message is a supervision message which is sent by node B96BN 170 to the host interface to indicate the availability of storage space in a master buffer pair 220 corresponding to a given destination terminal node. It is an upward flowing message and carries no parameter below byte position 3. Byte 0 contains 1, 0, 0, 1, 0 in bit positions 7-2 respectively, while byte position 2 carries 1, 1, 1, 0, 0, 0, 0, 0, in bit positions 7-0 respectively. Input supervision No. 36 is carried in byte position 3 with the node address indicating the terminal node address of the data stream to which the message applies.

The permission to send message is used in the message and block set motion feedback loop that exists between the host 12 and the large two port buffer memory 172 for messages going to a terminal node. The permission to send message originates in node B96BN 174 and flows upward to Y node 110 where it is duplicated and presented to the host 12 and to net master node 134. Net master node 134 normally discards the message. The message indicates to the host that the terminal specified in the address field has an available block set buffer in the large two port buffer 172 which can be used to receive another message block set (up to 16 large blocks) from the host. The host can respond with another block set for the specified terminal node if there are any data messages available to send. If node B96BN 170 fails to receive another block set within a fixed time out period (for example 16 sec) it resends the permission to send message. Node B96Bn 170 continues to repeat this action until the host responds with a message block set. Another group of messages establishes an error detection and control feedback loop through the line master node LM96N 174 between large two port buffer 172 and each destination terminal node connected in the subnetwork. Supervision messages move through the master up and down quenes 227, 228 independent of the data stream flow through master buffer pairs 220.

The block set acknowledgment and permission to send message is communicated from a terminal note to the line master node, LM96N 174, when a complete message has been received and the terminal node has buffer storage available for receipt of another message. Byte 0 bits 7-2 contain 1, 0, 0, 0, 1, and 0 respectively, while the node address field contains the address of the acknowledging terminal node. Byte position 2 contains in bit positions 7-0 respectively, 1, 1, 1, 0, 0, 0,0, 0. The input supervision type identified in byte position 3 is No. 32.

The ACKPRM message also contains 3 parameters in byte positions 4, 5, and 6. Byte position 4 carries a buffers available count (BACMF) which may indicate a maximum of 2 or only 1 buffer available in the present implementation for storage of a message from the line master LM96N 174. Each buffer has sufficient storage capacity for one message containing up to 16 large blocks or 24 bytes each. Byte position 5 carries an expected block sequence number (ESNMF) which indicates the expected block sequence number and would normally be 0 when the message is sent upon receipt of a completed message. Byte position 6 carries a block set sequence number (BSSNMF). This is a number that recycles between 0 and 255 and is incremented for each complete message or block set. The number indicates the number of complete block sets that have been received. This parameter provides error checking information to permit the line master to determine if a complete block set has been lost in the communication process.

Each terminal node has implemented therein a first empty cell pointer which is incremented only as complete blocks are received error free in numerical sequence. For example, while the pointer points to cell 8 of a message, a block 9 might be received followed by block 8 out of order. The pointer would remain at cell 8 until block 8 is received, at which time it would be rapidly incremented through cell 9, to cell 10 and remain at a count of 10 until block 10 is received. Upon incrementing of the first empty block cell pointer beyond No. 15 or beyond the number of a block containing an end of message value of 1 in the end of message signal field, message ACKPRM is sent by the destination terminal node if buffer storage space in available at the terminal node.

The sequence alarm or status request response message is sent from a dentination terminal node to line master node LM96N 174 to indicate the status of the buffers in the terminal node. It is identical in format to message ACKRPM except that the input supervision type number in byte position 3 is 33 instead of 32. Message ALMSRM is sent unsolicited if a block is received at a destination terminal node whose block sequence number is greater than the expected sequence number plus a value N. For example, it might be determined that in a given communication network a given block cannot get out of sequence during normal operation by more than 3 block positions. A number N might then be selected to be 3. If a block is then received carrying a block sequence number greater than or equal to the number indicated by the first empty cell pointer plus N equal 3, an error condition is indicated and the message ALMSRM is sent unsolicited. For example, if the first empty cell pointer contains No. 7, and the next block received contains block sequence No. 9, there would be not problem. However, if a block containing sequence No. 10 is then received before the first empty cell pointer is incremented to 8 upon receipt of sequential block No. 8, an error condition would be indicated and message ALMSRM would be sent. The first empty cell pointer field in byte position 5 would then contain the number of the first empty cell pointer to indicate to the line master the block within a message at which retransmission of the message should begin. The redundancy of retransmitting the first several blocks of a message which have been received error free is thus omitted while proper transmission of all blocks is assured even if an error condition develops.

The request status message is sent by line master LM96N 174 to a designated terminal node to determine the status of the terminal node. Byte position 0 carries 1, 0, 0, 1, 1, 0 in bit positions 7-2 respectively, and the terminal node address is carried in the node address field. Byte position 2 carries 1, 1, 1, 0, 0, 0, 0, 0, in bit positions 7-0 respectively, and byte positions 3 carries an output supervision type (OTSTSF) No. 40. No parameter is carried in byte positions 4-22 and these are all set to 0. Message REQSTM is employed by LM96N 174 in block set motion management from a buffer 220 to a buffer in a terminal node when node 174 needs to determine a terminal node status relative to block set motion. One usage examle is when line master node 174 has timed out a block set acknowledgment. Before taking any action, line master 174 determines the terminal node status by sending message REQSTM and receiving message ALMSRM in response thereto. Line master node 174 then has sufficient information available to continue the transmission of information using the status information received in message ALMSRM.

This message is identical to message ACKPRM except that the input supervision type number in byte position 3 is 34 instead of 32. This message is sent from a terminal node to the line master node LM96N 174 when the first empty cell pointer is incremented beyond a last cell in a buffer or beyond and end of message block of data and there is no block set buffer available in the terminal node into which a next block set can be received. Message SETAKM indicates to the linemaster node 174 that the destination terminal node has received a complete message or block set, but that no space is available at the terminal node for the receipt of an additional block set.

The above four error control feedback loop messages provide an elegant, yet simple and low overhead communication system to insure the error free communication of data through the error prone communication links. Error free communication requires very little hand shaking, yet if an error condition does occur, only a few superfluous data blocks are sent before the error condition is indicated to the line master and retransmission begins with the first incomplete data block to avoid the duplication of retransmitting data blocks that have been received error free. At the same time, programming at the terminal node is minimized by the relatively simple error checking, counting and storing process while the number of supervision messages is maintained quite small. It is also an advantage of the system that the supervision messages use the same information communication facilities as data messages and therefore require no separate or extra hardware. The substantial data storage in the buffers at the terminal nodes and in large two port 172 at the line master assure the continued and uninterrupted availability of data at each active terminal to permit smooth, uninterrupted data communication at a terminal itself even during short periods of maximum capacity network usage. The acknowledgement, alarm and status request messages assure that the buffer storage space is used to good advantage without buffer overflow and without an empty buffer going unusued when data is available for a destination terminal from the hose data processor 12 which communication capacity is also available from the network. While in the present arrangement, two 16 cell message buffers are implemented at each terminal node and at the line master for each terminal node, it will be appreciated that the number of buffers and the number of blocks for each buffer can be varied to optimize the communication network for any given application.

Three messages are available for controlling a full duplex communication link. As explained previously, whenever an uncorrectable error condition occurs along a full duplex data link, the link may be reconfigured as two simplex data links in such a manner that communication is maintained with a maximum number of terminals. The reconfiguration messages facilitate this procedure.

This message is generated by line master LM96N 174 and is communicated along a 9.6 KB line to one of the D96NX nodes such as node D96NW 260 or node D96NE 266. The message commands the designated node to configure its line exchange unit device in a loopback arrangement with messages received over on 9.6 kilobit line from the line master being sent back toward the line master on the coextending parallel data line. The lookback direction is not specified by the message but is inherent in the selection of the D96NX node and the hardware configuration of the line exchange unit device.

The data format uses 1, 0, 0, 1, 1, 0 in byte position 0 bits 7-2 respectively. The node address identifies the D96NX node within the line exchange unit device which controls the loopback function. Byte position 2 bits7-0 respectively, contain 1, 1, 1, 0, 0, 0, 0, 0. The output supervision type identified in byte position 3 in No. 38 and the loopback message contains no parameter in byte positions 4-22.

This message is a counterpart of the loopback message except that it contains all 1's in the node address field and commands all D96NX nodes along a communication link to configure the associated line exchange unit device in a non-loopback arrangement. This message carries 1, 0, 0, 1, 1, 0 in byte 0 bits 7-2 respectively, and 1, 1, 1, 0, 0, 0, 0, 0 in byte position 2 bits 7-0 respectively. The output supervision type number is 34. No parameters are carried in bit positions 4-22.

The loop continuity message is used by line master LM96N 174 after issuance of a loopback message to determine whether a loop in one direction or the other has a continuous path. The line master issues a series of these messages and tests for each return to the line master. If, over a series of 3 outputs, 3 messages are returned back to the line master, the line master assumes the loop in question has continuity. Line master LM96N 174 commands one of the OE96NX nodes such as OE96NW 184 to place the loop continuity message on the data link. The message contains the node address of the line master and moves through the data link of the network as if it were an upward flowing message from a terminal. The line master recognizes the continuity message as a supervision message does not pass it on toward the host data processor when it arrives back at the line master. The line master also periodically issues the loop continuity message normal operation, for example every second, to monitor the 9.6 KB loops for continuity to determine when to automatically reconfigure a full duplex line into two simplex facilities to isolate failed segments.

The message format includes 1, 0, 0, 0, 1, 0 in byte position 0, bit positions 7-2 respectively and 1,1,1,0,0,0,0, in byte position 2 bits 7-0 respectively. The input supervision type number in byte position 3 to 35 and no parameters are carried in byte positions 4-22.

The network control messages provide the operating control over the communication network. They permit the net-master node, operating in response to a control console or in response to the host data processor, to read the contents of any addressable storage location at any node in the network or to write into any writable addressable storage location at any node in the network. The addressable storage locations include the RAM and ROM memories within each node itself as well as the RAMs which provide the Q's to interconnect to nodes. It thus becomes possible to send both data and program information through the network for storage at a given node. Each node can then be commanded to execute a program to operate on the data. The communication system can thus be configured as a large, selectively programmable, multi-processor system with processing capabilities, program storage and data storage distributed throughout the substantial data storage and processing capacity of the network.

This message provides a capability of reading one, two, or three contiguous words (8 bit bytes) at any start address in any addressable node in the network. The message emanates from the host in response to a host program or from the network console. If the node address fields contain address Hex 001, the netmaster node 134 is addressed and otherwise the NCRDM message is relayed through the netmaster node to a designated node in the network. A destination node for the message responds by reading the designated address locations and issuing a node control response message (NCRSPM) in response thereto. This moves the contents of the specified word locations to the netmaster node 134.

Message NCRDM contains 1, 0, 0, 1, 1, 0 in byte position 0, bits 7-2 respectively. The node address is the destination node from which stored data is being read and byte position 2 contains 1, 1, 1, 0, 0, 0, 0, 0 in bit positions 7-0 respectively. The output supervision type number at byte position 3 is 33 and several parameters are carried in byte positions 4-10. Byte position 4 carries a read start address high field (RSAMF (HI)) to identify the 8 most significant bits of a start address while byte position 5 carries a read start address low field (RSAMF (LO)) to identify the 8 least significant bits of a start address. Byte position 6 contains a number of words field (NOWMF) to indicate whether the reading of one, two, or three 8 bit words starting at the read start address is being commanded. Byte positions 7-9 contain all 0's and byte position 10 contains a message content check sum field (MCCSMF) which provides a simple byte addition check sum over the message parameters in byte positions 0-9. The usual check sum for byte positions 0-22 is carried in byte position 23 in the large block form of the message. A destination node receiving message NCRDM checks both the message content check sum and the regular check sum over the entire message and simply discards and ignores the message if an error is detected in either.

The network control response message is issued by a designated node in response to the receipt of a network control read message NCRDM or a network control write message NCWRM in response to the net control read message, message NCRSPM carries the contents of the specified locations back to the netmaster 134 and then from there normally on to the host or to the control console. In response to a net control write message, message NCRSPM carries the contents of the specified locations (after the write operation) back to netmaster 134 and from there normally on to the host. It is utilized by the netmaster to verify a correct write operation.

Message NCRSPM carried in byte position 0 bits 7-2 respectively, 1, 0, 0, 0, 1, 0. The nodes address is that of the responding node and byte position 2 carries 1, 1, 1, 0, 0, 0, 0, 0 in bit positions 7-0 respectively. The input supervision type number in byte position 3 to 37 and the parameter content of the message is carried in byte positions 4-10. Byte positions 4 and 5 respectively, carry the most and least significant bytes of a 16 bit start read or start write address. Byte position 6 carries the number of words being read or written, which may be one, two or three. Byte positions 7-9 designated RCMFO-2 carry the first, second and third words as appropriate and byte position 10 carries a message content check sum field (MCCSMF) which is a simple byte addition over the byte positions 0-9. The usual message check sum over byte positions 0-22 is carried in byte position 23 in the long block form and the netmaster node discards and ignores the message in the event that an error is indicated by either of the check sums.

The network control write message commands the writing of up to 3 bytes of information into any addressable, writable address location at any selected node. Byte position 0, bits 7-2 contain 1, 0, 0, 1, 1, 0 respectively, and the node address identifies the designated addressable node. Byte position 2 contains 1, 1, 1, 0, 0, 0, 0, 0 and byte position 3 contains an output supervision type No. 35. Byte positions 4 and 5 contain fields (WSAMF (HI) and WSAMF (LO)) which respectively indicate the most and least significant bytes of a 16 bit address field. Byte position 6 indicates the number of words to be written in the field, NOWMF, which may vary from 1-3. Byte positions 7-9 contain fields WCMFO-2 with the up to 3 words that are to be written. Byte position 10 carries a message content check sum over byte positions 0-9 and the normal check sum over byte position 0-22 is contained in byte position 23 to indicate a simple byte addition check sum over byte positions 0-22. If an error is detected from either of the two check sums, the message is simply ignored and discarded by the destination node.

All call messages are issued by the netmaster node 134 on a regular periodic basis approximately every 22 seconds. The all call messages permit the netmaster node 134 to maintain in tabular storage a running account of the status of various parameters throughout the entire communication network. They also permit automatic reconfiguration of the communication network in response to call response messages. As the all call response message flows upward through the network, each node which receives the message and must make a decision about downward paths for destination messages, updates a table indicating the path over which the all call response message is received for each downstream responding node. The all call messages are also utilized as they pass around the communication links to determine the preferred link over which an upward flowing message will be sent when a choice between two different lines is available.

Byte position 0 bits 7-2 respectively contain 1, 0, 0, 1, 1, 0. The node address contains all 1's and byte position 2 contains 1, 1, 1, 0, 0, 0, 0, 0, in bit positions 7-0 respectively. Byte position 3 carries the output supervision type number which may be one of six numbers varying from 0 through 5. Each number illicits a different response parameter from a responding node in the all call response message therefrom to permit the netmaster node 134 to maintain a table of 6 double byte parameters for each node. Each time a new all call message is issued after a 22 second interval the output supervision type number is incremented so that after 6 all call messages over a period of approximately 132 seconds, the status table in the netmaster node 134 is completely updated. The parameters are as follows:

0. number of small down blocks reaching the node.

1. Number of large down blocks reaching the node.

2. Number of up blocks reaching the node.

3. Total number of block sets (messages) reaching the node.

4. Number of single small block type of block sets reaching the node.

5. Not used--all zeros.

0. number of empty buffer available signals sent to node B96BN 170 to command issuance of message PRSNDM.

1-5. not used--all zeros.

0. number of small down blocks reaching the node.

1. Number of large down blocks reaching the node.

2. Number of request status messages, REQSTM, reaching the node.

3. Number of abort messages, ABORTM, reaching the node.

4-5. Not used--all zeros.

0. number of up blocks reaching the node and not discarded.

1. Number of up blocks reaching the node, but discarded.

2. Number of down blocks reaching the node, all are discarded.

3. Number of alarm and status response messages, ALMSRM, reaching the node.

4. Number of block set acknowledge messages, SETAKM, reaching the node.

5. Number of acknowledge and permission to send messages reaching the node.

0. number of single small down blocks reaching the node and not discarded.

1. Number of large down blocks reaching the node and not discarded.

2. Number of down blocks discarded.

3. Number of up blocks reaching the node.

4-5. Not used--all zeros.

0. number of single small down blocks reaching node.

1. Number of large down blocks reaching node.

2-5. Not used--all zeros.

0. number of up blocks reaching node and not discarded.

1. Number of blocks reaching node, but discarded.

2. Number of down blocks reaching node--all are discarded.

3-5. Not used--all zeros.

0. number of small down blocks reaching node and not discarded.

1. Number of large down blocks reaching node and not discarded.

2. Number of down blocks discarded at node.

3. Number of up blocks reaching the node.

4. Number of alarm and status request messages, ALMSRM, reaching node.

5. Number of hesitations in the delivery of a data block to an associated terminal because of the unavailability of output data while an output data message is in the process of being received by the terminal node.

Byte position 4 contains a node position count message field (NPCMF) that is incremented by each of the line drop micronodes D96NW and D96NE as it passes along the 9.6 KB data links 196, 198. This field enables each of the line exchange devices 62 to learn its position on each line with respect to the line master exchange unit device 60. This information is then used by the CIEN 272 node to determine the preferred path for upward flowing data blocks. The node position count is also incremented by the terminal exchange unit devices and utilized to determine their position along a single 57.6 KB coaxial cable. Byte positions 5-22 contain all 0's and the usual check sum is contained in byte position 23 in the large block format.

The all call message is originated by netmaster node 134 via queues HCDQ 132 and NMONQ 138 every 22 seconds. The message is copied for distribution along all downward switch paths and is discarded at the input end nodes IE96NE 180, IE96NW 188 and CIEN 272. As the all call messages flows outward from the netmaster node it distributes timing and configuration control throughout the network. The node B96BN 170 uses this timing for deallocating a flip-flop buffer 220 from a terminal. That is, if an all call response message is not received within a selected time out period such as 176 seconds, node B96BN 170 assumes that the terminal has been disconnected and the buffer pair 220 assigned thereto is deallocated. Similarly, the netmaster node 134 times out the existence of nodes in its hierarchy data base if an all call response is not received after the generation of eight sequential all call messages.

The all call messages are delivered in synchronous parts of sub-networks as follows. Each message is taken off the 9.6 or 57.6 KB line into storage, processed and later relayed from storage via the 9.6 KB line or the 57.6 KB coaxial cable. All synchronous nodes along the coax including the terminal nodes, node CIEN 272 and node COEN 282 respond to the all call message with the all call response message after a period of time equal to 12 (A+8) F., where A is the address of the responding node and F. is the frame time in seconds. A frame time is the time required to send a short data block of 12 bytes or 96 bits along the associated data link. Nodes B96N 170, LM96N 174, the fan out nodes 54, 56, and all synchronous nodes along the 9.6 KB channels including D96NW 260, D96NE 266, IE96NE 180, IE96NW 188, and OE96NE 192 and OEN96NW 184 all respond after a time (2) (A+8) (F). The data rate is 1/6 as fast so the multiplier is decreased by a factor of 6 from 12 to 2 so that the response times are consistent and comparable for all nodes in the network whether they are along the high speed coax cable or along the slower 9.6 kilobit DDS channels. In most implementations, the YN 110 is not an addressable node and does not respond to the all call message.

The all call response message is generated by each of the addressable terminals in response to an all call message after a time delay proportional to the node address. The 0 byte position of the all call response message contains 1, 0, 0, 0, 1, 0 in bit positions 7-2 respectively. The node address field contains all 1's and byte position 2 contains 1, 1, 1, 0, 0, 0, 0, 0 in bit positions 7-0 respectively. The input supervision type number in byte position 2 reflects the output supervision type number of the all call message to which the response message relates and thus varies between 0 and 5. Byte position 4 contains a responding node position count field (RNPCMF) which is completed by nodes D96NW 260 and D96NE 266. These nodes retain the incrementing count field of the original all call message as it passes through the 9.6 kilobit line and insert it in their response message. This enables the netmaster to determine the position of the node in the network and the line master to determine the position of the node in the subnetwork.

Byte position 5 bit positions 7-4 contain a node type field designated NTYPMF with 0 not being used, 1 indicating a terminal node, 2 indicating a CIEN node, 3 indicating a COEN node, a 4 indicating a D96NE or D96NW node, a 5 indicating an IE96NE or IE96NW node, a 6 indicating an OE96NE or OE96NW node, a 7 indicating an LM96M node, and an 8 indicating a BN96BN node. An output rate value field RNORMF is implemented with bits 7 and 6 of the field contained in bit positions 3 and 2 of byte position 5 and with bit positions 5-0 contained in byte position 7, bit positions 7-2 respectively. Terminal nodes insert into this field the bit rate at which data is transferred to a connected terminal. This information is utilized indirectly by the line master node to select the dedicated buffer pair 220 from which a next block of data will be sent through the network to a destination terminal node. Nonterminal nodes insert all 0's into this field. A responding node address field (RNAMF) is implemented at byte position 5 bit position 1 and 0 and at byte position 6, bit positions 7-0. This field contains the address of the node generating the all call response message. Similarly, an upper node address field (UMAMF) in implemented at byte position 7, bit positions 1 and 0 and in byte positions 8, bit positions 7-0. This field contains the address of the node next above the responding node in the network hierarchy. The responding node places all 0's in this field and any node which receives an all call response message containing all 0's in this field places its own address in the field.

This information is utilized by the netmaster node 134 to maintain an updated mapping of the communication network structure on a dynamic basis notwithstanding reconfiguration of the network from time to time by removing existing terminals or adding terminals. Byte positions 9 and 10 contain a 2 byte parameter for the responding node statistic (RNSTMF) which varies with the type of node and with the six types of output supervision type number of the original all call message which is being responded to. 0's are inserted in byte positions 11-22 and the conventional check sum is inserted in byte position 23 in the large block format.

The Y node 110 duplicates all all call response messages and presents them to the netmaster node 134 and to the host adapter. This is part of the Y node action on all upward flowing messages and the all call response messages are discarded by the host adapter. The B96BN node 170 detects the number 1 indicating a terminal node in the NTYPMF field and responds by allocating a buffer pair 220 corresponding to the address field of the responding terminal node. The line master node LM96N 174 utilizes the responding node position count field for nodes along the 9.6 KB line to obtain information on node distribution along the 9.6 KB lines and maintain a node table for reconfiguration purposes. The LM96N 174 node uses field RNIEMF to identify the IE96NE or IE96NW node response defining the end of the reconfiguration node table.

The reinitialization message is issued by the line master node LM96N 174 and causes the node to which it is addressed to execute its initialization program and start operation with a fresh RAM load in which all earlier data and status information is discarded. This message is discarded by the destination node in the event of a check sum error.

Byte position 0 bits 7-2 contain 1, 0, 0, 1, 1, 0 respectively. The address of the destination node is inserted in the node address field and byte position 2 bit positions 7-0 are set to 1, 1, 1, 0, 0, 0, 0, 0. The output supervision type number is contained in byte position 3 and is 41 for the reinitialization message. Bytes 4-9 contain all 0's and byte 10 contains a redundant check sum. Bytes 11-22 for the large block format contain all 0's and byte position 23 contains the conventional check sum.

These messages enable a network operator to obtain at a console terminal a display of the data base stored in netmaster node 134 (FIG. 2) which relate to the structure and operation of the communication network. A display data base message (DSBM) is generated in response to an operator command, and flows outward through the network to the netmaster node 134. The netmaster node responds to the display data base message by generating a one of a plurality of different types of response messages which communicate a desired portion of the network data base back to the originating terminal.

If the operator request is entered through the TTY control console, the display data base message is communicated from console interface node 148 upward through queue LCUQ 146 to the netmaster node 134 and the response is communicated back through LCDQ 145 to console interface node 138. On the other hand, if the display data base message originates at another terminal in the network or with the host data processor 12, the message reaches the net master node 134 through HCUQ 131 and the response is returned to the host through HCDQ 132. The netmaster node 134 maintains paired associations between each addressable node in the network and the node immediately thereabove in the hierarchy as well as 6 double byte parameters for each node. The display data base messages may command the response to provide the entire data base or selected portions thereof which depend upon the input supervision type number for the display data base message and particular parameters that are contained in the message. Each input supervision type number commands a specific response message and the requested information will be made clearer in the following discussion of the response messages. In the present implementation of the system 10, the programming has not been provided for the network control interface messages because of their redundancy to the network control messages which are used instead. However, the interface messages are somewhat more convenient to use and the required programming can readily be provided by a person of ordinary skill in the art to accomplish the functions described below.

The format of the display data base message includes at byte position 0, bit positions 7-2 respectively, 1, 0, 0, 0, 1, 0. The node address associated with a requesting terminal is placed in the node address field at byte position 0 bits 1 and 0 and at byte position 1, bits 7-0. This would typically be the address of the console interface node 148 but might also be the address for any terminal node in the network. Byte position 2 contains 1, 1, 1, 0, 0, 0, 0, 0, in bit positions 7-0 respectively. The input supervision type number appears in byte position 3 and may vary from 16 through 26 in accordance with a desired response as indicated below. Byte position 4 contains a number x which varies between 0 and 5 to identify one of the 6 double byte parameters which are stored in the netmaster node 134 data base for the addressable nodes in response to the all call message. The parameter x may or may not have any significance for a given input supervision type number and is set to 0 if it has no significance. Byte position 5, bits 1 and 0 in combination with byte position 6 carry a Y parameter which is a specific node address for which data base information is being requested. This Y parameter may or may not have significance depending upon the particular input supervision type number and is set to 0 if it does not have any significance. Byte position 7, bits 1 and 0 in combination with byte position 8 define a z parameter which is a node address of a B96BN node such as node 170. This address defines a subnet that is implemented below the given node in the hierarchial structure. In the present implementation there is only one subnet, but in general a communication network may contain several subnets and the parameter defines a particular subnet of interest. In the large block form, 0's are entered in byte positions 9-22 and the check sum is entered in byte position 23.

The basic message format for all of the display data base response messages is the same and includes at byte position 0 bits 7-2 respectively, 1, 0, x, 1, 1, 0. Bit 5 is the block type identification field and its contents depend upon the particular data content of the block. A 1 indicates that it must be transformed into a double small block and a 0 indicates that it may be transformed into a single small block for communication along the serial data links. In a multi-block response message, only the last block might possibly be transformable into a single small block and that would depend upon the number of data bytes in the block. The node address field reflects the node address contained in the originating display date base message. Byte position 2 bits 7 and 6 contain 1, 1 respectively. Bit position 5 is the end of message signal field and contains logic 1 for the last block of a message and logic 0 for preceding blocks of a multi-block message. The block sequence number in bit positions 4-0 sequentially number the blocks of a response message from 0-15. Byte position 3 carries the output supervision type number which reflects the input supervision type number of the display data base message to which the netmaster node 134 is responding. Similarly, byte position 4 reflects the X parameter, byte positions 5 and 6 reflect the Y parameter, and byte positions 7 and 8 reflect the Z parameter of the originating display data base message. A conventional check sum appears in byte position 23 of each data block and the intermediate data positions carry the data information in the form stored in the netmaster node 134 data base. This data will vary widely with the particular configuration of the network and with the input supervision type number of the originating display data base message.

The all hierarchy pair response message (AHPRM) is generated in response to input supervision type number 24 and carries the hierarchy pair for each node in each subnetwork as identified by the B96BN node and all nodes therebelow. The hierarchy pairs are the paired associations of a given node address and the address of the node immediately thereabove in the network hierarchy. The parameters X, Y and Z have no significance for input supervision type number 24.

The all hierarchy and statistic pair response message (AHSPRM) is generated in response to input supervision type number 25 and contains the entire netmaster node network data base. The parameters X, Y, and Z have no significance in this message.

The all subnet statistic pair response message (ASNPRM) is generated in response to input supervision type number 19 and carries 1 of 6 double byte statistics for all of the nodes of all the subnets as identified by the parameter X. X corresponds to the all call message type numbers for the data desired and thus varies between 0 and 5. The parameters Y and Z have no significance.

The all statistics pair message (ASPRM) is generated in response to input supervision type number 21 and carries the double byte statistics data for all 6 parameters associated with every addressable node in the network. The parameters X, Y and Z have no significance.

The node all statistics pair response message (NASPRM) carries the double byte statistics for each of the 6 parameters associated with a specific addressable node identified by parameter Y. The parameters X and Z have no significance.

The node hierarchy chain response message, (NHCHRM), is generated in response to input supervision type number 26 and carries the hierarchy pair for all nodes upward of a node whose address is specified by the parameter Y. The parameters X and Z have no significance.

The node hierarchy pair response message (NHPRM) is generated in response to an input supervision type number 22 and carries the hierarchy upper node pair for a specific node whose address is specified by the parameter Y. The parameters X and Z have no significance.

The node statistics pair response message (NSPRM) is generated in response to input supervision type number 16 and carries the double byte statistical parameter specified by parameter X for a specific node whose address is specified by parameter Y. Parameter Z has no significance in this response message.

The subnet all statistics pair response message (SNASRM) is generated in response to input supervision type Number 18 and carries all statistics pairs for all 6 parameters for a particular subnet identified by the address of the subnet B96BN node carried by parameter Z. Parameters X and Y have no significance for this response message.

The subnet hierarchy pair response message (SNHPRM) is generated in response to input supervision type number 23 and causes the netmaster node 134 to provide all hierarchy pairs for all addressable nodes above a given B96BN node in the network specified whose address is specified by the parameter Z. Parameters X and Y have no significance in this message.

The subnet statistics pair response message (SNSPRM) is generated in response to input supervision type number 17 and carries the double byte statistic identified by parameter X for each node in a specific subnet identified by parameter Z. The parameter Y has no significance in this response message.

The network control interface messages thus enable a person operating the communication network to rapidly obtain on a real time basis the present status of the dynamically changing communication network configuration as well as the statistical parameters relating to network operation. For convenience, the entire statistical data base may be obtained or the data base may be obtained for only specified portions of the network.

The patentable novelty of this invention is deemed to lie in the disclosed method and apparatus for providing data communication rather than in the specific programs which are provided for the various programmable nodes in the system. Furthermore, the specific program load listings for each node are provided at the end of this Detailed Description. An exception is the programming for fan out nodes 54 and 56. Up to the date of filing the application for this patent alternate path 90 has been used and the programming for the fan out nodes 54, 56 has not been completed. However, it will be appreciated that the programming for the fan out nodes can be provided in accordance with the teachings herein. It is thus deemed unnecessary to provide detailed flow diagrams for each of the operating programs of the system. However, to promote a better understanding of the general operation of the communication system, a generalized discussion of system functions and programming will be provided first as to functions and programming which relate to the system generally or to a plurality of nodes in the system and then as to specific nodes.

Each programmable node is built around a separate microprocessor and has its own independent internal programming and address structure. Each node has a ring of high level programs which are executed in a sequential, recirculating manner. Each high level program relates to a different primary function such as the inputing of data from a queue or from a data channel, the outputing of data to a channel as channel capacity becomes available, maintanance of internal timing functions, and resetting of a hardware time out for implementation of a deadman's stick. Outside of the high level program ring are an initialization program and perhaps an interrupt program. The initialization program is entered only in response to a command external to the node, received either over a hardware control line or as a supervision message received as any other message flowing through the network. The initialization program is exited to one of the high level programs where recirculation through the ring commences. An interrupt program is entered in response to an interrupt request and after execution of the interrupt program, a return is made to the next instruction in a sequence of instructions that was being executed at the time of the interrupt. The high level programs may include subroutines to perform certain functions or subprograms, but the high level program exit is always to the next high level program in the ring upon completion of a high level program task. For example, the high level program task might be to sample an input such as a queue in a two port buffer memory and perform a specified operation on any block of data that appears in the input. Thus, as the high level program is executed, the input is sampled and if no data is found, an exit is made to the next high level program in the ring. On the other hand, if data is found, control might be transferred to a subprogram which tests for whether or not the data block relates to a data message passing through the network or a supervision message. Control would then be further passed to another subprogram depending upon the type of message that is being input until the proper high level program function is performed and then control would be passed to the next high level program in the ring. Subroutines may be used within any of the programs or subprograms.

Queues are used for transfer of blocks of information through two port buffer memories between nodes and for transfer of 12 byte small blocks or blockettes between programs in the same node. The communicating programs sharing a queue within a node may be on the same program level or at different levels. For example, an interrupt program may insert or remove data from a channel input/output module along a data link while a high level program positioned within the ring examines the information, performs any necessary operations, and either receives the information from a connecting two port queue or passes the information on through a connecting two port queue. Each cell of a queue is said to store a large block or small block of data and a queue may be read from or written into without affecting the status of the queue if the pointers are not changed. However, a block of data is said to be placed in a queue when the next write pointer is incremented and is said to be removed from the queue when the next read pointer is incremented. After writing into the cell or reading from the cell, the appropriate pointer is incremented and the block is then said to be placed in the queue or removed from the queue.

As explained previously, each queue has associated therewith a plurality of cells, one or two address maps, and a single control package. The cells and control packages are located in random access storage and, for a queue in a two port buffer memory, this storage is part of the memory. The address map is normally located in read only memory and if two nodes share a connecting two port, each node has an address map for the queues of the two port.

Each terminal node is provided with a buffer pair, and a buffer pair is implemented in the large two port memory of the net master node of a subnet for each terminal node. Each of the buffer pairs consists of two standard queues with 16 24 byte cells each. Each queue is thus capable of holding a maximum of on full output data message.

The output data messages are loaded into and removed from the buffer pairs in rotary. Current buffer write and current buffer read fields associated with each buffer are maintained by the sending and receiving programs. If the current write field contains 0, buffer 0 of the pair is being written into by the sending programs and if it contains a 1 then buffer 1 is being written into. Similarly, if the current read field contains a 0, the receiving program is reading from buffer 0 and if it contains a 1, the receiving program is reading from buffer 1. The buffer pair is empty when both of its queues are empty. Space is available for an additional output data message if one or both of the queues are empty. The 32 buffer pairs in the large two port memory between the B96BN node 170 and the LM96N node 174 can of course be increased in number if necessary to accommodate a larger number of terminal nodes in the subnetwork. The output data messages are moved under control of the permission to send message issued by the B96BN node 170 from the host data processor to a queue in a buffer pair of the large two port memory 172 on a per stream basis. That is, a complete, uninterrupted message is sent from the host to a queue in a buffer pair upon receipt of a permission to send message when data is to be communicated. The output messages are then moved on a per stream basis from a large two port buffer to a corresponding terminal node buffer. The data is moved on a block by block basis with the blocks being selected by a line allocation algorithm which has been explained previously.

The 24 byte large blocks of data are converted to either a single or double 12 byte small block of data for communication along the data links. Identical programs are used for block conversion and blockette motion on both the 9.6 kilobit and 57.6 kilobit loops. Therefore, in the following discussion, the term drop node refers to the D96NE, D96NW and TN nodes. The term input end node refers to the IE96NE, IE96NW and CIEN nodes and the term output end nodes refers to the OE96NE, OE96NW and COEN nodes. A small block arriving at a drop node along a channel or data link may be carried either by a single or a double small block. A double small block is used if the net data content of the block is greater than 8 bytes. In the present implementation, only type B or downward flowing messages are large enough to require a double small block, with all supervision blocks and all upward flowing blocks being represented by a single small block. Thus, all upward flowing blocks which are presented to a drop node for delivery to the channel contain 8 bytes or less of net data and are therefore transformed into a single small block before being stored in a queue for output on a channel.

Upon the arrival of the first double byte of a block of data at a drop node, the drop node utilizes the node address to access an internal table to determine whether the block should be allowed to pass along the data link or whether it should be pulled off.

An empty frame arriving at a drop node presents an opportunity to output a small block of data to the channel. This opportunity is exercised by filling the frame with a small block waiting in a queue. Similarly, a small block being switched off a channel at a drop node creates an empty output frame and presents an opportunity for the node to output a small block of data. If an arriving frame contains a small block of data and another small block of data is waiting in a queue for outputing, a priority decision must be made to determine whether the incoming small block stays in its frame and continues along the channel or whether it is to be diverted to a small block "siding" queue in internal drop node storage and replaced by the waiting upward flowing small block. The priority relationships have been established with the first or highest priority being given to an A (upward flowing) small block arriving at the drop node in a data frame along the communication link. The second priority is given to type A small blocks that have arrived or been generated at a node other than via the communication link and is awaiting output. The third priority is assigned to type B or downward flowing blocks which are waiting in a siding gueue for output. These are small blocks which have been previously pulled off the data link to make room for higher priority upward flowing blocks. The lowest priority is assigned to type B downdward flowing blocks which arrive at the drop node along the channel.

A drop node is interrupted on arrival of each double byte of data from the input channel. At the time of arrival, the hardward presents to predetermined node address locations the last 16 bits that have arrived over the channel. The hardware also presents a D byte count (0-5) at a third address location. This information is held until the next interrupt. When the double byte count is equal to 1, the second double byte is being serially entered and the hardware is presenting the first of six double bytes in parallel to the associated micronode. This is the double byte which contains header and address information. The drop node may thus examine the first double byte, and within a single double byte delay time, make a decision to pass the block along the communication link, pull the block off the communication link or substitute another block to be passed along the communication link.

The drop node program maintains an internal double byte counter which is incremented by one on a modulo six recirculating basis at each interrupt. Upon interrupt, the internal double byte is compared with the hardware double byte counter. If the two counters are equal the program continues with its normal actions. If the two counters are not equal, the program enters a loop that merely examines the hardware double byte counter and stays in this loop until the hardware counter becomes equal to the program counter. While in this loop input data from the channel is allowed to flow through the input shift register to the channel output shift register and be output on the channel. When the two counter fields become equal, the program proceeds with its normal activity. Any resulting errors due to lost double bytes are corrected by the error loop control between the LM96N node and the destination terminal nodes.

On arrival of the last double byte in a frame at a micronode as indicated by D byte count No. 0, a drop node program prepares a summary of the state of the node for use on arrival of the first dual byte of the next frame. In preparing this summary, the program selects the highest priority gueue (if any) having a small block available for output in the next frame and stages the oldest blockette of this priority in the output queue for possible output during the next frame. If no small block is available for sending, the program stages an empty frame for possible output. Any previously staged but unsent small block is returned to its queue before the new summary is prepared.

On arrival of the first double byte in the next frame, a decision is made as to whether the arriving or staged small block has highest priority and whether the address of the arriving small block specifies that it should be switched off the channel at the drop node or continued along the channel. Depending upon the priority and node address conditions, an arriving small block may be moved to the siding gueue or a local receiving gueue for small blocks being switched off the channel or may be allowed to flow through and continue along the channel. An empty frame arriving from the channel is replaced by a staged small block, which might also be an empty frame.

The drop node hardware provides a pair of bus address locations into which the program may write for output on the channel. If, after one interrupt and before the next interrupt, the program writes a double byte into these two locations, this double byte will replace the double byte that has arrived from the input channel. Thus, for a swap, the drop node program moves two bytes from the input double byte bus address locations to a cell of a selected gueue and moves two bytes from the staged small block to the output double byte address locations at each of the six interrupts which represent a frame carrying a small block of data. This process begins with interrupt No. 0 while the D byte count is 1. For outputing data only, the drop node program moves double bytes from the staged small block to the output double byte locations at each of the six interrupts representing a frame, but does not read from the input locations. If a data frame and the small block it contains are to pass through the drop node, the program takes no action in moving double bytes to or from the hardware during the six interrupts of a small block.

For moving small blocks of data from the channel to an input end node, the programming is a subset of the programming that is used at a drop node. Since an input end node cannot place data on a data link, the input portion of the program may be deleted. The input end node accepts small blocks from the channel and places them in a small block arrival queue. If, at interrupt No. 0, the incoming frame is found to be empty, the double bytes 0-5 are allowed to move through the shift register of the associated input channel and are lost. If, at interrupt 0, the incoming frame is found to be not empty, a double byte is moved at each of the interrupts 0-5 to the next write cell of the small block arrival queue.

Similarly, an output end node has no capability of removing data from a channel and the programming for an output end node is essentially a subset of the drop node programming. An output end node accepts small blocks of data from an output small block queue and places them on the channel in synchronism with the data frames. If no small block is available from the output queue, the output end node places an empty frame on the channel. As with the drop node, either a small block or an empty frame is staged in the output queue at interrupt number 5 for delivery of a double byte to the channel during each of the following six interrupts 0-5.

The deadman's stick is implemented to provide automatic reinitialization of a programmable micronode in the event that program execution thereat goes errant. For example, a bit error might cause the microprocessor to improperly interpret an instruction and almost anything can happen as a result. Since such errors may occur from time to time without any actual malfunction, it is possible that proper operation of the node can be restored merely by reinitialization. This is accomplished with a hardware one shot which is periodically reset by writing into a selected bus address location. The program for the deadman's stick is one of the high level programs in the program ring and is periodically entered to reset the one shot during normal operation. However, if a program error occurs, the deadman's stick program will not be executed and the one shot will time out. The timing out of the one shot causes a hardware commanded initialization with the microprocessor being constrained to jump to an address location defined by the hardware implementation of the microprocessor. The microprocessor than proceeds to execute the initialization program and return to normal operation.

During initialization the random access memory of each node is set to initial values by the initialization program that is resident in the node's read only memory. The initialization can be commanded by either a pushbutton switch or by the time out of the one shot of the node's deadman's stick. All nodes at a given exchange unit that are connected by two port buffer memories are commanded to start initialization procedures simultaneously upon the time out of the deadman's stick or upon reset pushbutton activation at any of the connected nodes.

A time out is implemented by a count down routine following each initialization and precedes hand over of the processor to the normal function defined by the high level program ring. The individual time counts for each of the programs of two port connected nodes are adjusted so that the time elapse between start of initialization and start of normal operation is approximately the same for all nodes in the interconnected set. For ease of implementation, a time period of one second, which is longer than is required for initialization of any of the nodes, is selected as the standard initialization period. This guarantees that initialization is complete in the multinode environment before a normal operation is commenced by any one node. To prevent redundancy and interference, the rather arbitrary rule is used that each node in the network initializes any two port buffer memory below it in the tree structure hierarchy as well as any two port buffer memory that is not connected to any other node.

Program interrupts are implemented only at nodes NMN, OE96NE, OE96NW, D96NE, D96NW, IE96Ne, IE96NW, COEN, TN, and CIEN. The application of the interrupt signal to the microprocessor at a node causes the control to be passed to an interrupt vector location to start the interrupt routine. On completion of the interrupt routine, control is passed back to the interrupted program at the next instruction by a return from interrupt instruction executed by the interrupt routine as its last step. Only one interrupt level is used and the programs installed in a given node are either high level or an interrupt program which is itself not interruptable.

The source and implementation of network timing at the various nodes depends upon the location of the node within the network. At the nodes which are connected to the data link, each node is interrupted six times for each frame time period to provide the necessary internal timing. This is six times for each 10 milliseconds for nodes on the 9.6KB channel and six times for each 1.667 millisecond period for nodes on the 57.6KB channel. This is sufficient to meet internal node timing requirements.

The netmaster node requires timing to produce the all call messages at the scheduled 22 second intervals. This timing is provided by the netmaster interrupt module which is controlled to generate an interrupt every 22 milliseconds.

The line master node requires timing at a 10 millisecond rate to schedule its activities relative to frame availability on the two 9.6KB output channels serviced by the OE96NE and OE96NW modes. It receives the 10 millisecond timing redundantly from both of these nodes via the connecting two ports as follows. The OE96NE and OE96NW nodes each set a timing field in the two port connecting unit to the LM96N node to logic 1 every 10 milliseconds, using the channel interrupt timing as a source. The LM96N node contains a field which, if set to 0 specifies the OE96NE node as the source of timing, and if set to 1 specifies the OE96NW node as the source of timing. If the source pointer points at the OE96NE node, the timing field in the two port maintained by the OE96NE node is tested. If it has the value 1, a clock tick for the LM96N node is generated and the shared timing field is reset to 0. If it has the value 0, no action is taken unless it has had the value 0 for the latest n samples. In this case the source pointer field is complemented and the OE96NW node is specified as the timing source.

Sevice of the current timing field is accomplished by the program in the LM96N program set ring. By setting n large enough, a crude timeout (much greater than 10 milliseconds) of the timing source is obtained. When this time out fires the alternate timing source is accessed. Thus, failure of either one of the output channels will not deprive the LM96N 174 node of the timing it needs to service the remaining channel. Clocks maintained by the LM96N 174 node in the large two ports shared with the B96BN node provide a time out capability to the B96BN node 170.

A processor input matching node (PIMN) 104 receives input data messages, permission to send messages, the net control response messages and all call response messages. On receipt of an input data message, INDATM, from the Y node, the processor input matching node 104 constructs a data packet in the host format using the originating node address and the data from the input daa message. The new packet is communicated to the host through the matching module 100. All messages except the input data message and the permission to send message arriving from the Y node 110 are discarded with no action taken by PIMN 104. On receipt of a permission to send message, PRSNDM, from the Y node 110, node PIMN 104 constructs a permission to send packet in the host format containing the originating node address. This packet is sent to the host.

Upon receipt of a net control response message, NCRSPM, from the node NMIFN 108 through NTRQ 119, PIMN 104 constructs a supervision packet in the host format. This packet contains the net control response message and is passed to the host.

The processor output matching node, POMN 105, receives all of the messages flowing downward from the host through processor output interface module 102. Upon receipt of an ABORT packet in the host format, an ABORTM message in the network format containing the address of the destination terminal node is generated and passed to Y node 110. Output data messages, OTDATM, are received in the host format from the host terminal and converted to an output data message of 1 to 16 blocks and passed on to Y node 110. The longest output data message is 16 blocks times 160 bits per block equal 2,560 bits. Longer bit streams are converted into multiple messages. All blocks except the last block of an output data message are loaded with exactly 160 bits of data and identified as a large block which must be moved through the serial data channels as a double small block. If the last block of the output data message is loaded with 64 bits or less it is marked for conversion to a single small block and is otherwise marked for conversion to a double small block.

Upon receipt of a net control read or net control write packet from the host, node POMN 105 converts to the message to the proper NCRDM or NCWRM network format containing the address of the destination node and passes the message to the netmaster interface node, NMIFM 108.

The netmaster inferface node (NMIFN) 108 receives the net control response messages, NCRSPM and passes them upward from the POMN 105 to netmaster node 134. It also receives the net control read and write messages NCRDM and NCWRM and passes them up from the POMN 105 to netmaster node 108.

The netmaster node (NMN) 134 receives the input data messages, the permission to send messages, the network control response messages, the all call response messages, the net control read messages, the net control write messages and originates the all call messages. The input data messages, INDATM, and permission to send messages, PRSNDM, arriving at NMN 134 from Y node 110 are simply discarded without action.

A net control response message received by the netmaster node 134 from the Y node is passed to the netmaster interface node 108 or to the console interface node 148 depending upon whether network control resides in the host or in the TTY control console. The location of network control is defined by the contents of a field in the netmaster node 134. A net control response message generated at NMN 134 in response to a net control read or write message contains the contents of the storage locations specified in the requesting message (after the requested write, if any). The net control response message generated within NMN 134 is passed to the netmaster interface node or the CIFN node 148 depending upon whether control resides in the host or the TTY console. A net control read or write message arriving from the host via the netmaster interface node 108 transfers control to the host. A net control read or write message arriving from the TTY control console via the console interface node 148 transfers control to the TTY control console.

Upon receving an all call response message, ALCLRM, the netmaster node 134 places the contents of the all call response message parameter in a data base storage area associated with the node originating the response message, as defined by the node address of the message. This storage area associated with each responding node is divided into a plurality of fields storing various information parameters pertaining to each responding node. One of six node statistics fields is updated with a corresponding one of six possible parameters which is carried by the all call response message. The parameter depends upon the all call message type number of the message which generated the response and a sequential rotation through all six types occurs. A responding node type number is stored in a responding node type field. The upper node address inserted by the node immediately above the responding node in the tree structure is placed in one of two upper node address fields. Two fields are needed because there may be two nodes above a given node (D96NE and D96NW are above CIEN). If both the responding node address and the upper node address fields are empty (both contain 0) the upper node address is stored in field No. 1. If only one field is empty, the upper node address contained in the response message is compared with the contents of the non-empty field. If the comparison results in an equality, no action is taken. If the upper node address in the message is different from the non-empty stored field the upper node address from the message is placed in the empty field. If neither of the two address fields of the all call response message is empty, the upper node address of the message is compared with the contents of each of the two stored address fields. If the upper node address is equal to the contents of either field, no action is taken. If the upper node address is not equal to the contents of one of the fields, both fields are emptied (set to zero) and the upper node address is stored in field No. 1. For nodes located along the serial data channel, the node position count is stored in a node position count field. After the netmaster node data base has been updated with the contents of the all call response message, the all call response message is discarded.

An all call message, ALCALM, is generated by the netmaster node 134 approximately every 22 seconds and passed to Y node 110 for distribution to all nodes below. The six different types of all call messages carrying the supervision message type numbers 0-5 are issued in a rotary sequence. The all call message type numbers specified to the destination nodes elicit six statistic fields in the receiving node which are loaded into the all call response messages by a responding node. The all call response message generated at a responding node carries the supervision message type number of the all call message which is being responded to. If a net control read or write message, NCRDM or NCWRM, contains the address of netmaster node 134, the requested read or write is performed and the results are used ro generate a net control response message. The net control read or write message is discarded. If the net control read or write message arrives from the netmaster interface node, the network control field in the netmaster node 134 is set to point to the host. If it arrives from the console interface node 148, the net control field is set to point to the TTY control console. If the net control read or write message does not contain the address of the netmaster node, the message is passed to the Y node 110 for distribution to the network and the network control field in the netmaster 134 is set to point in the direction from which the message came.

The console interface node (CIFN) 148 handles net control read messages, net control write messages, and net control response messages. If a net control response message is received by the console interface node 148 from the netmaster node 134 in response to a net control read message, the contents of the net control read message are displayed on the TTY network console. If the requesting net control read message is not the last of a request sequence, the next net control read message is generated. If it is the last, the sequence is terminated. If a net control response message is received by console interface node 148 from the netmaster node in response to a net control write message, the contents of the net control response message are used to confirm successful completion of the write specified by the net control write message. The TTY network control console supplies the contents for the net control wrire messages sequentially with two hexidecimal characters for a single byte or four characters for a double byte write or six characters for a three byte write. The arrival of the necessary characters from the console released the net control write message.

A read command from the TTY network control console results in the console interface node 148 generating a sequence of one or more net control read messages as necessary to read the storage area specified as a start location and length in the read command. The read command requires actuation of key "R" followed by a 3digit hex node address followed by a 4digit hex internal start address followed by a 4 digit hex byte length number. A write command is similar except key "W" is depressed instead of key "R" and instead of a byte length, pairs of hex characters for sequential bytes are supplied. The write continues sequentially until a non-hex key is actuated to end the operation.. The first net control read message of the read sequence is passed on to the netmaster node on an unsolicited basis. The remaining net control read messages of the sequence are issued to the netmaster node 134 upon receipt from the netmaster node of the net control response message corresponding to the previously sent net control read message. Bytes of data are read in groups of two using the net control read message until a specified number of bytes has been read.

The TTY network control console supplies the write parameters for net control write messages six hexidecimal characters at a time (or less for the last write message). As the console interface node 148 receives each set of six hex characters it releases a net control write message to the netmaster node 134. Netmaster node 134 executes the net control response message if the write message is addressed to the netmaster node 134. Otherwise, netmaster node 134 passes the net control write message on to the Y node 110 for distribution to the network. If the network control console supplies only one or two hex characters prior to terminating the write command with a carriage return, a single byte net control write message will be generated. If three or four hex characters are supplied a two byte net control write message will be generated and if five or six hex characters are supplied a three byte net control write message will be generated. One, two or three bytes may thus be written into a node at a specified bus address location within a specified node.

The Y node (YN) 110 responds to receipt of input data messages, INDATAM, and permission to send messages, PRSNDM, by duplicating the messages and sending them both to the netmaster node 134 and to the processor interface matching node 104. The net control response messages received from below are also duplicated and sent to PIMN 104 and NMN 134. In the event that Y node 110 is addressable (not in the present implementation) and generates a net control response message, the internally generated response message is also duplicated and sent to PIMN 104 and NMN 134. The Y node generates an all call response message only if it is implemented as an addressable node. Similarly, the Y node 110 places its address in an empty upper node address field only if it has an address assigned thereto. All all call response messages, whether internally generated or received from below are duplicated and communicated to the NMN 134 and PIMN 104. Abort messages received by YN 110 from POMN 105 are passed through to the network below. Similarly, output data messages are received from POMN 105 and passed through to the network below. Net control read and write messages received from NMN 134 are passed through to the network below except when YN 110 is addressable and the message contains the Y node address. In this case, the control message is discarded after execution and an appropriate net control response message is generated and communicated to NMN 134.

The fan out nodes 54 and 56 merely pass through network messages that are not addressed to them. They make the usual response to call call messages and generate network control response messages in response to net control read and write messages addressed to them. They maintain a table by responding node address of the path below on which a node responds and use the table to direct downward flowing messages over the proper downward path.

Node B96BN 170 receives input data messages, INDATM, passing upward through the network from MUQ 228 and communicates them through the fan out nodes to the Y node.

The permission to send message, PRSNDM, is normally generated by the mode B96BN 170. The message contains the node address of a terminal node for which a buffer store is available in large two port 172 and is sent upward through MUQ 228 to the host 112. As an output data message is sent to a terminal node and an empty buffer becomes available upon receipt of a terminal node acknowledgement message, node LM96N 174 provides a signal through large two port 172 and the node B96BN assembles and sends a permission to send message upward to the host. Node B96BN also maintains a time out counter of 16 seconds plus a period of time proportional to the node address of a terminal node to which a data stream having an empty buffer applies. When this time out expires, a new permission to send message is sent to the host. It is quite possible that 16 seconds can pass without data being sent to a terminal node in the normal course of operation. However, the passing of this time could also mean that an error has occurred and the host data processor has failed to get the permission to send a message. The time out assures that such an error is self-correcting after a short period of time.

The net control read and write messages are normally passed down to the network through MDQ 227. If the message is addressed to node B96BN 170 a proper net control response message is prepared and sent upward through the network. Similarly, net control response messages generated below node B96BN 170 are merely passed upward through the node 170.

Node B96BN 170 responds to an all call message with an all call response message in a manner similar to all other nodes. In addition, if an all call response message received from node LM96N 174 below contains an empty upper node address field, node B96BN 170 inserts its own address into this upper node address field. Otherwise, all call messages received from below are merely passed upward through the network. However, if an all call response message received from below identifies its source as a terminal node, node B96BN either creates a data stream for that terminal node (by assigning a buffer pair to the terminal node) or verifies the existence of a data stream for the responding terminal node. Node B96BN 170 maintains a node address table field associated with the address of the originating terminal node. This field indicates whether the terminal node is active with a data stream or buffer pair assigned thereto or is inactive. If eight all call messages pass through the node B96BN 170 without receipt of an all call response message for a given stream, the stream is deactivated. Upon receipt by node B96BN 170 of an abort message, ABORTM, from above, the message is merely passed downwardly through MDQ 227 to node LM96N 174.

The output data message arrives at node B96BN 170 from the host processor above one block at a time with the last block of the message containing a logic 1 in the end of message field. Node B96BN 170 stores the blocks of a message in the current write buffer or a buffer pair assigned to the destination terminal node as indicated by the address field in the output data message block. If, upon storage of the last block of a message an empty buffer of the corresponding buffer pair is available, node B96BN 170 generates a permission to send message associated with the data stream. The node address table field of the node B96BN 170 is utilized to provide an association between the terminal node address of an output data message and the corresponding buffer pair.

Upon receipt of an all call message from above, node B96BN 170 passes the message onward through MDQ 227 and generates its own appropriate response.

Line master node (LM96N) 174 operates to selectively send data messages from the buffer pairs downward through the subnetwork to a terminal node on a per data stream basis. The stream with which a message is associated is identified by the terminal node address contained in the message.

The line master node 174 maintains error loop control by maintaining a parameter for each data stream, the parameter identifies one of six possible states for the data stream. State 0 is the output enable state. In this state the blocks of an output data message are moved from the current read buffer of a buffer pair in large two port 172 through the subnet to a terminal node below. The streams that are in the output enabled state are available for selection of a next block of data using the line allocation algorithm as explained above.

State No. 1 is an acknowledgement expected state. This state is entered after a last block of data of an output data message is sent to a destination terminal node. In this state, the line master node 174 awaits an acknowledgement and permission to send, ACKPRM, or block set acknowledgement, SETAKM, message from the terminal node corresponding to the data stream. As the data stream enters state 1, a time out is set and if this time out period expires without receipt of an acknowledgement message the data stream is changed to state 3. Upon receipt of an acknowledgement and permission to send message the state reverts to state 0 and upon receipt of a block set acknowledgement message the state changes to state 2.

State 2 is the space expected state. In this state a complete message has been sent and a block set acknowledgement response message has been received from the destination terminal node. This means that both buffers at the terminal node are full and the data stream awaits receipt of an acknowledgement and permission to send message, which indicates that a buffer has become available to receive additional data. Upon receipt of the acknowledgement and permission to send message, control passes to state 0. Upon entering state 2, a time out period is set and if this period passes without receipt of the acknowledgement and permission to send, control passes to state 4.

State 3 is the no acknowledgement abnormal state. The stream enters this state from state 1 as a result of a failure to receive an acknowledgement either in the form of an acknowledgement and permission to send message or a block acknowledgement message during a time out period initiated upon output of the last block of a message. Upon entering this state, status is requested and a time out is initiated. Each time a time out expires a new status request message is sent and a time out is again initiated. The data stream remains in this state until the data stream becomes inactive or until an acknowledgement message transfers control to state 0 or state 1.

State 4 is a no space abnormal state. The stream enters this state from a state 2 as a result of failure to receive an acknowledgement and permission to send message from the associated terminal node during the state 2 time out period. Upon entering state 4, a time out is set and a status request message is sent to the terminal node. Each time the time out expires, a new status request message is sent and the time out is reset. The data stream remains in state 4 until a status response message is received which indicates that control should be transferreed elsewhere.

Stage 5 is an ABORT state. The data stream enters this state upon receipt of an ABORTM from above and remains in this state until an acknowledgement with permission to send message is received from the terminal node. The acknowledgement with permission to send message indicates that the terminal node has responded to the abort message by clearing its buffer storage. Upon entering state 5 a time out period is set to count eight data frame periods and upon expiration of the time out a new abort message is sent and the time out is reset. Upon receipt of the acknowledgement with permission to send message control is transferred to state 0.

Node LM96N 174 periodically sends a loop continuity message, LPCNTM, around each of the halves of the duplex loop and if the message fails to return three times in succession for a given loop, a break in the communication link is presumed to exist. The line master node 174 then proceeds to execute the isolation and correction program as follows. Line master node 174 issues a loopback message, LPBAKM, along the east loop addressed to the D96NE drop node at the farthest end of the east loop. After a pause to allow the loopback message to be received and executed, the line master node issues a loop continuity message, LPCNTM, to the OE96NE node 192 node for delivery to the east loop output and examines the queue from the IE96NW node 188 to determine if the loop continuity message is returned to it through the reconfigured loop as previously commanded through the last D96NE node along the loop. If the loop continuity message returns to the line master node within a specified time interval the loopback is left in place and the reconfiguration procedure goes to the west loop. If the loop continuity message does not return within the specified time, a clear loopback message, CLPBKM, is issued on the east loop, followed by a loopback message, LPBAKM, to the next to the last D96NE drop node on the east loop. After another pause for execution of these commands, another loop continuity message is sent to test the reconfigured loop continuity. This procedure is followed on the east loop sending the loopback message to successively closer D96NE drop nodes within the line exchange units until continuity is established or until all of the D96NE drop nodes on the east loop are exhausted. At this time the reconfiguration procedure starts with the west loop proceeding in exactly the same way from the farthest line exchange unit from the input end toward the nearest unitl either continuity is established between the west output end node (OE96NW 184) and the east input end node (IE96NE 180) or until the attempt to establish has been made at each D96NW drop node and all attempts have failed.

If continuity is established from either east output to west input or from west output to east input or both, reconfiguration is a success and is left in place. If no continuity has been established, a clear loopback message, CLPBKM, is issued to each east and west drop nodes along both the east and west loops and another attempt is made to operate in the normal configuration. Once a reconfigured loopback continuity is established, it continues until the defect has been repaired and a net control write message causes a return to the normal state by setting a field in the line master 174 which causes one of the line master 174 programs to clear the loopback. In the event that the line master node is unable to establish any normal or reconfigured loopback continuity it continues to alternately try normal and loopback nodes until one is successful.

After completion of a successful reconfiguration loopback, within 44 seconds the next all call response messages result in an updating of the network hierarchy data maintained in the netmaster node 134 data base to reflect the change in the subnetwork configuration. The network configuration displayed on the network console is thus changed automatically in response to the reconfiguration and no special messages to a network maintenance operator are required.

Line master node 174 merely passes the input data messages, INDATM, through MUQ 228 to node B96BN 170 above. Upon receipt of an acknowledgement and permission to send message, ACKPRM, the line master node acts in accordance with the state of the associated data stream. In the acknowledgement expected state 1, the space expected state 2 or the abort state 5, stream control is passed to output enabled state 0, which enables the transmission of the next block set for this data stream to the associated terminal node. If an acknowledgement with permission to send message is received and the associated stream is not in one of these three states, the message is discarded and no action is taken.

Line master 174 receives the sequence alarm or status request response message, ALMSRM, from a terminal node when a block of data is received too far out of order or in response to a status request to the terminal node. If the data stream is in the output enabled state 0, the current output data block set is resent, starting with the block sequence number specified in the message. If the associated data stream is in the acknowledgement expected state 1, the space expected state 2 or the abort state 5, the message ALMSRM is discarded and no action is taken. If the data stream is in the no acknowledgement abnormal state 3, the action depends upon parameters contained in the ALMSRM message. If the output data message block sequence number contained in the ALMSRM message agrees with the output data message block sequence number stored for this stream in the line master node and space is available in the terminal node, transmission of the next data block of the output data message is enabled and the stream goes to the output enabled state 0.

If the output data message block sequence number contained in the ALMSRM message agrees with the message block sequence number stored for this data stream in the LM96N node, and no space is available in the terminal node according to the content of the ALMSRM message, the stream goes to the space expected state 2 and the usual timeout is initiated.

If the output data message block sequence number contained in the ALMSRM message does not agree, with the message block sequence number stored by the line master node, the current output data message block sequence number for the data stream in the line master node is reset to correspond to that contained in the ALMSRM message.

If the ALMSRM message is received while the data stream is in the no space abnormal state 4, and the ALMSRM message indicates that space is available at the terminal node, the stream goes to the output enabled state 0 which enables the transmission of the next output data message block set along this stream. If the ALMSRM message is received while the data stream is in the no space abnormal state 4 and the message indicates that space is not available at the terminal node, a request status message is sent to the terminal node and a timeout is initiated.

If a block set acknowledge message is received and the associated data stream is in the acknowledge expected state 1, the stream goes to the space expected state 2 and a timeout is initiated. If the timeout fires while the data stream is in the space expected state 2 without the arrival of an acknowledgement and permission to send message from the terminal node to indicate the availability of space at the node, a request status message is issued to the terminal node and the stream goes to the no space abnormal state 4 and a timeout is initiated. If a block set acknowledgement message is received in any state other than the acknowledgement expected state the message is discarded and no action is taken.

The network control read message, network control write message and network control response message are handled as at other nodes in the network. If the read or write message is addressed to the line master node, the appropriate action is taken and a response message is sent upward. Otherwise, the messages are passed upward or downward through the node with no action being taken.

The loop continuity message, LPCNTM, is issued and received by line master node 174 to detect continuity on either of the two simplex loops (east or west) comprising the 9.6 KB duplex loop. A loop continuity message carrying the line master node address is moved down through EDQE 191 to node OE96NE 192 once per second for delivery to the east simplex loop. Under normal conditions with no break in the east simplex loop the message will move along the loop to node IE96NE 180 where it is pulled off and transferred up through EUQE 179 to line master node 174. If timeouts maintained by the line master node 174 indicate that three sequential loop continuity messages fail to return, the line master node 174 determines that continuity has been lost on the east simplex loop and the reconfiguration procedure is initiated.

Line master node 174 similarly issues the loop continuity message once per second to the west simplex loop to test for continuity in exactly the same way as on the east simplex loop. Detection of a continuity failure in either of the east or the west simplex loops causes initiation of the reconfiguration procedure. The loop continuity message is also used to test for continuity of trial loops commanded by the line master node using the loopback message during execution of a reconfiguration procedure.

The line master node 174 responds to an all call message by preparing an all call response message and sending the message after a time delay proportional to the node address as is done by other nodes. As all call response messages are received from below, they are passed upward through the network. Messages from nodes IE96NW and IE96NE, which carry no upper node address, are operated upon by placing the address of the line master node in the upper node address field before passing the messages upward. If an all call response message is received which originated at a node along the east 9.6 KB channel including node IE96NE, a D96NE or node OE96NE, the line master node places the node addresses from the message in order in an east node sequential position table. Similarly, if the message originated at a drop node along the west 9.6 KB channel, the node address from the message is placed in order corresponding to the order of the node along the channel in a west node sequential position table within internal line master node storage. The east and west node position tables are used in reconfiguring the duplex loop in the event of a channel or node failure.

The abort message, ABORTM, is received from the B96BM node 170 above and is passed on to the addressed terminal node below. Upon receipt of an abort message, line master 174 initiates a timeout of 150 milliseconds and causes the associated data stream to go to the abort state 5. During the 150 milliseconds after initiation of the timeout and prior to the end of the timeout, the associated data stream is nonresponsive to all messages arriving from the terminal node. Also, upon receipt of the abort message, the line master node 174 empties the corresponding buffer pair associated with the stream, initializes the stream and signals the B96BN node 170 through the large two port 172 to issue a permission to send message to the host. This enables the sending of the next output data message by the host if there is any to be sent. At the end of the 150 millisecond timeout period, the data stream is again made responsive to responses from the terminal node, the abort message is passed on to the terminal node below, and a timeout is initiated. The first timeout enables the data stream path to be cleared before the abort message is sent to the terminal node. If the second timeout expires without receipt of an acknowledgement with permission to send message from the terminal node, another abort message is issued to the terminal node and the second timeout is again initiated. After eight occurrences of this timeout without a proper response, the stream is initialized at the line master node and a reinitialization message, REINM, is sent to the terminal node.

The abort message is passed down to the OE96NE or OE96NW node for eventual delivery to the addressed terminal node. Selection of the proper output node is based on the contents of the field in the node address table associated with the destination node for the abort message. Upon receipt of the acknowledgement with permission to send message during the second timeout, control is passed from abort state 5 to the output enabled state 0.

As space becomes available in the queues EDQW 183 and EDQE 191 leading to the inputs to the west and east simplex channel loops respectively, the line master node 174 selects an output message block from the output enabled data streams or from the downward flowing supervision path through MDQ 227. The downward flowing supervision blocks of data are given priority. The east or west output end node removes the selected block from the end node down queue, provides a conversion from a large block to a single or double small block format and inserts the small blocks into data time frames as they become available on the serial communication channel. The converted small blocks are placed in internal queue storage at the input end nodes while waiting for delivery to the channel. The queue control package for the private storage queue is located in the two port buffer memory 182 or 190 to make the contol information available to the line master node 174. The decision as to which, if either of the two port queues leading to the output end nodes east and west and output data message block should be placed is made as follows. If the private queue in node OE96NE contains less than three small blocks, the next block is placed in the two port queue 190 leading to node OE96NE. Otherwise, the private queue in OE96NW is examined by the line master node and if this queue contains less than three small blocks, the next block is placed in the two port queue 182 leading to the west output node. If neither internal queue contains two or less small blocks, all blocks of data are retained in the buffer pair storage in the large two port memory 172.

Each output data block is examined as it is moved from the current output buffer in the large two port buffer memory 172 to determine whether it is marked as a last block of the output data message or not. If it is a last block the stream is set to the acknowledge expected state 1. This inhibits the further transfer of blocks for that data stream.

Since the output capability represented by the two queues for the output end nodes is shared by 32 data streams corresponding to 32 terminal nodes, a line allocation algorithm is executed by the line master node 174 to determine which block of output data is to be selected next for transfer to a terminal node. The line master node estimates the number of blocks waiting in the terminal node for delivery to the associated terminal for each of the 32 terminals serviced by the line master node 174. The number of blocks of data stored at each terminal node is estimated by incrementing a per stream counter as each block is output along that stream and decrementing the counter at a fixed rate corresponding to the rate at which blocks are moved from the terminal node to the terminal. This fixed rate is obtained by the line master node 174 from the terminal node all call response messages.

The LM96N node 174 selects for output from the subset of nonempty buffer pairs corresponding to a state 0 data stream, the buffer pair having the least number of estimated blocks in the terminal node awaiting delivery to the terminal. In this way the line master node 174 operates to maintain equal amounts of data in reserve for delivery to the terminals from the local storage at the terminal node.

The request status message, REQSTM, is issued by the line master node 174 to obtain a status response message from a designated terminal node. In the acknowledgement expected state 1 the request status message is sent when a timeout expires for the data stream. The timeout indicates that the terminal node has failed to receive a complete output data message block set or that there has been a loss of the acknowledgement with permission to send or block set acknowledgement message. In space expected state 2 the request status message is sent at the expiration of the timeout, which indicates that an acknowledgement with permission to send message has not been received because space is not yet available in the terminal node or the acknowledgement message has been lost. In no acknowledgement abnormal state 3 the status request message is sent at the expiration of the timeout. This timeout indicates that a previously requested alarm and status response message has not been received from the terminal node. When a data stream is in the no space abnormal state 4 the request status message is sent at the expiration of the timeout period. This timeout indicates that a previously requested alarm or status response message has not been received from the terminal node. The status message is also sent when the data stream is in the no space abnormal state for an alarm or status response message is received which indicates that space is still not available at the terminal node. The status request message is not issued in states 0 and 5.

The loopback message, LPBAKM, is issued by the line master node 174 to command a serial data channel reconfiguration on detection of a channel failure. This reconfiguration technique has been previously described. The clear loopback message, CLPBKM, is issued by the line master node 174 to either the east or the west simplex data channel. This message is not addressed to a specific node but instead to all nodes along the channel by using the all 1's all call address in the address field. It is delivered to all input end nodes, output end nodes and drop nodes along the given east or west serial data channel and causes any drop node receiving it to remove an existing loopback.

The line master node 174 initiates a reinitialization message, REINM, for a given data stream under specific circumstances. A timeout counter is maintained in the line master node for each stream. This counter is incremented at each timeout initiation and is set to 0 when a response is received from the terminal node. If this counter has the value 8 when a timeout occurs, the following actions are substituted for the specified normal actions. A reinitialization message is sent to the terminal node for the data stream and after a 1.5 second delay, the output data message in the current buffer of the line master buffer pair for the data stream is discarded and transmission of the next output data message block set (if any) is enabled.

The output end nodes OE96NE 192 and OE96NW 184 contain identical programs and function in an identical manner, except for being connected to a different east or west serial data channel. They respond to the net control read or write messages and the all call messages as any other node in the system. The messages are passed through or responded to with an appropriate net control response message or all call response message. Other messages are received from the line master node 174 and passed along the connected serial data channel. The messages are received in a large block form and converted to an appropriate single or double small block form before communication along the channel. The output end nodes are part of a unidirectional loop with all downward flowing messages being received through EDQE 191 or EDQW 183 and all upward flowing messages passing through the channel. In the hierarchy sequence, the output end nodes are deemed to be at the lowest end of the hierarchy with all of the drop nodes and finally the input end nodes being above the output end nodes in the hierarchy. Upon receipt of an abort message, ABORTM, a request status message, REQSTM, a loopback message, LPBAKM, a cancel loopback message, CLPBAKM, a loop continuity message, LPCNTM, or a reinitialize message, REINM, from the line master above, the output end nodes generate a check sum for the block containing the message and provide the conversion of the block to a small block format, and transmit the small block along the channel as data frames become available. Similarly, upon receipt of an output data message block from the line master node above, a check sum is added by the output end node, an appropriate conversion to a small block format is made and the single or double small block of data is output on the serial data link as data time frames become available. The output end nodes also add a check sum and convert the all call messages to the small block format. The all call response messages for the output end nodes are generated in the proper small block format with check sum attached.

The input end nodes IE96NE 180 and IE96NW 188 pull the data frames off of the serial communication channels and convert single or double small blocks of data to a large block format and communicate the large blocks to the line master node 174 through the end up queues EUQE 179 and EUQW 187. Before communicating a block of data to the line master node, the check sum is tested and the block is discarded if an error has occurred during the course of transmission. The input end nodes respond to the net control read, net control write and all call messages as do other nodes in the network. The input end nodes simply discard the messages output data message, OTDATM; request status message, REQSTM; loopback command message, LPBAKM; cancel loopback message, CLPBKM; and reinitialize message, REINM. The all call response messages that are received from the channel as a single small block are converted to a 24 byte large block format and the check sum is tested. If the check sum is invalid the all call response message is discarded with no further action being taken.

The drop nodes D96NE 266 and D96NW 280 interface the 9.6 KB serial data channels with the end nodes for the coaxial cable serial data channels.

An input data message, INDATM; acknowledge and permission to send message, ACKPRM; alarm and status response message, ALMSRM; or block set acknowledge message, SETAKM, if received by a drop node along the channel, is merely left on the 9.6 KB channel and passed on. If the message is received by the drop node from the coax input end node, CIEN 272, it is output on the 9.6 KB channel as any other message. A loop continuity message, LPCNTM, arriving at a drop node along the 9.6 KB channel is merely left to continue along the channel. The net control read, net control write and all call messages are responded to as at other nodes with the net control messages being directed toward an addressed node and the all call messages being duplicated and both continued along the channel and output toward the coaxial cable. Responses to these messages are prepared and communicated along the 9.6 KB channel as appropriate. If an all call response message is received from the 57.6 KB coaxial cable channel, the check sum is tested and the message is communicated along the 9.6 KB channel if valid. Otherwise the all call response message is discarded with no action. Similarly, as an all call message is received, it is converted to a 24 byte large block format and the check sum is tested. If the check sum is faulty the all call message is discarded without duplication. If valid the message is sent to the coaxial loop below and the node position count is incremented before a duplicate of the message is passed along the 9.6 KB line.

The coax output end node (COEN) 242 receives data from the two east and west drop nodes and outputs the data on the coaxial cable. It responds to the net control messages and the all call messages as other nodes in the network. The node COEN receives no all call response message from any other node and passes its own all call response message onto the first terminal node along the coax loop. The abort message, ABORTM; request status message, REQSTM; and reinitialize message, REINM; are output along the coaxial cable as received from one of the two connected drop nodes. The coax output end node receives data blocks from the drop nodes in the 24 byte large block format with the blocks being transformed into the appropriate small block format. Output data messages, OUTDATM, may thus be transformed into either single or double small blocks for transmission on the coax data channel.

The coax input end node (CIEN) 272 receives blocks of data from the coax serial data channel in small block format, provides a transformation to the large block format and communicates them to one of the two east or west drop nodes. The usual response is made to the net control and all call messages as appropriate.

As the input data message, INDATM; acknowledge and permission to send message, ACKPRM; alarm or status response message, ALMSRM; or block set acknowledge message, SETAKM, are received by the coax input end node, the check sum is tested and the block is passed to one of the drop nodes if valid. The block is discarded if the check sum is invalid. Upward flowing messages are delivered to the drop node having the fewest intermediate drop nodes between it and the output end node for the 9.6 KB serial data channel. If the number of intermediate drop nodes is equal for both the east and the west channels, the east drop node, D96NE is arbitrarily chosen. An abort message, ABORTM; request status message, REQSTM; reinitialize message, REINM, or output data message OUTDATM is discarded with no action if received from the coax channel.

Each terminal node (TN) 316 forms the downstream end of one of the data stream error control loops for the subnet controlled by the line master node 174. The terminal node receives an output data message, ODTATM, from the coax channel a block at a time. The blocks of a given message may or may not arrive in the order of their block sequence numbers. Each block is placed in the cell of the current input buffer of a buffer pair implemented in internal terminal node storage having a cell number equal to the block sequence number contained in the output data message block. The terminal node maintains several pointers for the buffer pair including a first empty cell pointer which points to the first cell in the current buffer in numerical sequence that has not received a block of data from the line master node. Even if a block is received out of sequence, the first empty cell pointer is not incremented until a block is received for the cell that is pointed to.

The terminal node also maintains a next read pointer pointing at the next cell from which a block is to be obtained to be passed on to the terminal matching node 324. When the next read pointer is equal to the first empty cell pointer the buffer pair is deemed to be empty and no data is available for output to the terminal node.

When the terminal node receives a block having a sequence number exceeding the value of the first empty cell pointer by a value of 4, an alarm and status response message, ALMSRM, containing the value of the first empty cell pointer is output on the channel for delivery to the line master node 174. If the first empty cell pointer is advanced to a value pointing to a nonexisting cell beyond the sixteenth cell in the current input buffer of the terminal node and an empty buffer of the buffer pair is available to accept another output data message, the terminal node outputs an acknowledgement and permission to send message on the channel for delivery to the line master node 174. If an empty buffer is not available, the terminal node outputs a block set acknowledge message SETAKN, on the channeel for delivery to he line master node 174. The terminal node than tests the buffer pair at short intervals for availability of an empty buffer. When this test indicates an empty buffer is available, the terminal node outputs an acknowledge and permission to send message on the coax channel for delivery to the line master node 174.

If the terminal node 316 receives a request status message, REQSTM, along the coaxial channel it responds by outputting an alarm and status response message along the channel. This message contains the value of the first empty cell pointer, the value of the stored block set sequence number and an indication of the availability of empty buffers. If the terminal node receives a reinitialize message, REINM, from the coaxial channel, the random access memory of the terminal node is loaded with initial values from a program stored in read only memory. If the terminal node receives an abort message, ABORTM, from the coaxial channel it blinds itself and discards all output data messages, OTDATM, request status messages, REQSTM, and reinitialize messages, REINM, for 0.15 seconds. It then discards the data in both buffers of the buffer pair and outputs an acknowledge and permission to send message along the coax channel for delivery to the line master node 174.

Any message received by the terminal node from the input coaxial channel having an invalid check sum is discarded with no action.

The current output buffer is examined periodically by the terminal node to determine if any blocks are available for transmission to the terminal matching node 324 which connects the terminal node 316 to a terminal. If a block is available and if there are less than two blocks in the queue TDWBQ 320 connecting the terminal node 316 to the terminal matching node 324, a block is transferred from the current output buffer to TDWBQ 320 leading to the terminal matching node 324.

An input data message, INDATM, received by a terminal node 316 from the terminal matching node 324 through TUDAQ 322 is output on the coaxial channel for eventual delivery to the host.

If upon receipt of a block of an output data message the first empty cell pointer is updated to a value pointing to a cell one number beyond a cell containing an output data message block marked as a last block (after a last block of an output data message is received) and an empty buffer of the terminal node buffer pair is available to accept another output data message, the terminal node outputs an acknowledgement and permission to send message on the output channel for delivery to the line master node. If an empty buffer is not available a block set acknowledge message, SETAKM, is sent to the line master node. After output of a block set acknowledge message by the terminal node 316, the terminal node 316 tests the buffer periodically for availability of an empty buffer. When this test indicates an empty buffer is available, the terminal node outputs an acknowledgement and permission to send message on the coax channel for delivery to the line master node 174. On receipt of a block of the output data message having a block sequence number exceeding the value of the first empty cell pointer by a count of four or more, an alarm status response message, ALMSRM, is output on the coax channel for delivery to the line master node 174. This message notifies the line master node 174 that an output data message block has been lost on the way down. On receipt of a request status message, REQSTM, along the coaxial channel, an alarm status response message is also output on the coaxial channel for delivery to the line master node 174. The alarm and status response message contains the value of the first empty cell pointer, the value of the stored block set sequence number, and the availability status of an empty buffer. The block set sequence number is initially set to 0 and is incremented upon receipt of each block set or message. The number is incremented to a total of 255 and then overflows back to 0. The block set sequence number permits the line master 174 to determine whether any complete messages have been lost during transmission.

If on receipt of a block of an output data message the first empty cell pointer is updated to a value pointing to a cell one count beyond a cell containing a last block of a message and an empty buffer of the terminal node buffer pair is not available to accept another output data message, the terminal node outputs a block set acknowledge message, SETAKM, on the output coaxial channel for delivery to the line master node 174. Upon receipt of a net control read or net control write message, the terminal node 316 performs the required operation and generates an appropriate net control response message.

Upon receipt of the all call message, the terminal node 316 prepares an all call response message containing the appropriate one of six parameters as identified by the type of all call message.

As all call response message received along the channel by a terminal node 316 is checked for a valid check sum. If invalid, it is discarded and if valid it is queued for output on the coaxial channel.

Upon receipt of an abort message, ABORTM, from the coax channel, a terminal node discards with no action all incoming output data message blocks, request status messages and reinitialize messages for 0.15 seconds. At the end of the 0.15 seconds the terminal node 316 discards all data in both buffers of the buffer pair and outputs an acknowledgement and permission to send message along the coax channel for delivery to the line master node 174. The terminal node 316 accepts only abort messages containing the terminal nodes address and leaves all other abort messages on the coax channel. All abort messages removed from the channel and having an invalid check sum are discarded with no action.

A terminal node 316 receives an output data message, ODTATM, from the coax channel a block at a time. If the block check sum is invalid the block is discarded with no action. If the block check sum is valid the block is placed in the cell of the current input buffer having a cell number equal to the block sequence number contained in the output data message block and the first empty cell pointer is updated. The terminal node 316 accepts only output data message blocks containing the address of the terminal node with all other output message blocks being left on the channel.

On receipt of a request status message, REQSTM, along the coax channel, that is addressed to the terminal node 316, the terminal node 316 responds by outputting an alarm and status response message along the coax channel. A request status message having an incorrect check sum is discarded. Request status messages not containing the terminal node address are left on the channel.

Upon receipt of a reinitialization message along the coax channel, the random access memory of the terminal node is loaded with initial values from the read only memory. Only reinitialization messages containing the terminal node address are input to the terminal node with others being left on the channel. Reinitialization messages addressed to the terminal node 316 but containing invalid check sums are discarded with no action.

The terminal matching node (TMN) 324 moves data in both directions between the terminal node 316 and the terminal console 328 through the terminal interface module 326. It also supplies a parameter to the terminal node 316 via the two port buffer memory 74 specifying the rate at which the terminal accepts data for display. This parameter is carried by an all call response message generated at the terminal node 316 and carried to the line master node 174 for use in estimating the amount of stored data at the terminal node 316.

The terminal matching node 324 accepts data from the terminal up to a maximum of 64 bits per block, places the data in an input data message, INDATM, and delivers the input data message to the queue TUADQ 322 in two port 74 for communication to the terminal node 316.

The terminal matching node 324 also accepts output data message blocks from the terminal node 316 through queue TDWBQ 320 in two port 74 and delivers the data content of the output data message blocks to the terminal console 328. Each block may contain up to 160 bits of data.

Although the particular selection of input/output and other address locations is in general of no special significance, the program load listings have been provided for a particular hardware connection arrangement. The following program address maps show these hardware connections.

The following is a list by node of the data stored at each node in read only memory. The four digit lefthand column identifies in hexidecimal notation an address location and the 16 two digit columns to the right thereof each identify in hexidecimal notation the contents of the 16 sequential address locations beginning with the address location indicated in the column to the far left. For example, in the netmaster node, address location F800 stores F9, address location F801 stores AE up to address location F8FF stores D2. The programming is for the instruction set of an MC6800 microprocessor, which is commercially available. ##SPC1## ##SPC2## ##SPC3## ##SPC4## ##SPC5## ##SPC6## ##SPC7## ##SPC8## ##SPC9## ##SPC10##

The following program load listings for nodes NMIFN, CIFN, YN and TMN have a slightly different format. The four left most digits of each line contain a 4 digit word pertaining to an assembler program which generated the load listings. The next 4 digits contain a beginning hexidecimal address location and the next 16 pairs of hexidecimal characters define the contents of 16 sequential address locations beginning with the defined address and the last pair of characters are of no interest. Some corrections or "patches" have been made in these load lists subsequent to their printout. The corrections for each node follow the node load list. ##SPC11## ##SPC12##

While there have been shown and described above a particular example of a communication system and distributed multiprocessor system in accordance with the invention for the purpose of enabling a person of ordinary skill in the art to make and use the invention, it will be appreciated that the invention is not limited to the particular example. Accordingly, any modifications, variations or equivalent arrangements within the scope of the attached claims should be considered to be within the scope of the invention.