Data Packet Format

This application is related in subject matter to the following applications filed concurrently herewith and assigned to a common assignee:

Application Ser. No. 07/636,795 filed by Nagy entitled "Improved Error Recovery Protocol".

Application Ser. No. 07/636,794 filed by Nagy entitled "Early Frame Release In Modified HDLC Protocol".

The foregoing co-pending applications are incorporated herein by reference.

1. Field of the Invention

This invention relates to a method of transmitting data frames, and more particularly, to a protocol and data frame packaging format optimized for transmissions over a communications link.

2. Background of the Invention

The need for economical high speed transmissions of data over communication links continues to increase. Transmission efficiency falls and costs rise drastically when data is required to be retransmitted. The possibility of having to retransmit data increases in noisy environments, such as transmissions over telephone lines using modems. Under the popular Synchronous Data Link Control (SDLC) protocol , when a data frame in a sequence of frames is lost, the protocol recovery rules force the re-transmission not only of the frame which was damaged, but all subsequent frames as well.

A better solution in the form of the High-Level Data Link Control (HDLC) protocol stipulated by the IOC (International Organization for Standardization) has been used to ease the problem. In the HDLC protocol, the REJ (Reject) scheme and the SREJ (Selective Reject) scheme are stipulated as re-transmission schemes to be used in the case where transmitted frames are in error and have not properly reached the receiver.

The REJ scheme is representative of the so-called "Go-Back-N Automatic Repeat Request scheme." The REJ scheme works in the following manner. When a receiver detects the omission of a frame because of the reception of an out of sequence frame, it sends back a signal REJn (where n represents the number of the erroneous frame) indicating a re-transmission request. The sending side retransmits the erroneous frame and also sends a series of sent frames succeeding thereto. The REJ scheme has the advantage that a small number of buffers suffice on both the sending and receiving sides.

According to the SREJ scheme, only erroneous frames are retransmitted. Therefore, no frame is transferred wastefully. However, two or more signals SREJ cannot be simultaneously sent back according to provisions of the HDLC protocol. Therefore, the SREJ capability experiences appreciable degradation in especially noisy conditions where multiple errors occur in a single burst of frames. Under such conditions, the SREJ capability reverts to the SDLC re-transmission rule, namely, the erroneous frame as well as all subsequent frames. Large numbers of buffers are also required on both the sending and receiving sides. These buffers serve to hold the error-free frames until the error frame is received and is properly sequenced.

HDLC frames must be released from the receiving link station in the exact order in which they are delivered to the transmitting link station. In noisy environment where retransmissions are taking place, this means that frames which have been correctly received cannot be released to their destinations until other unrelated packets have been retransmitted without error. The problem revolves around a fundamental provision of the protocol, namely, that HDLC assumes responsibility for re-sequencing the frames, to maintain data synchronization. The provision applies regardless of whether the receiver has an intelligent application layer capable of re-sequencing the frames.

Consequently, what is needed is a protocol and data packaging format which provides for the simultaneous identification and selective re-transmission of a plurality of erroneous data frames received in a single burst of data frames, while releasing error-free data frames immediately to their destinations.

This invention relates to a protocol and data packaging format optimized for transmissions over noisy communications links. Full duplex operation is provided over communications channels where transmission errors are expected to be encountered. The control mechanism is incorporated into a normal data communication header thus eliminating the requirement for control messages to acknowledge or reject transmissions. This invention provides an eight bit value which encodes a toggle bit status for each of eight possible transmit sequence numbers. The eight bit value when used alone, constitutes a null frame indicating data-free feedback. This null frame is used to confirm receipt of data frames. This eight bit value is also used as the first byte of every data frame. The second byte of a data frame consists of a four-bit logical channel index, a single bit transmit toggle bit and a three-bit transmit sequence number. The second byte of the data frame may be followed by zero or more eight-bit data bytes. When a data frame arrives at a receiver, the bit value of the transmit toggle bit is copied into one of the eight bits of the first byte according to the transmit sequence number. The receiver sends the newly modified first byte back to the sender as part of a data frame transfer or as a null frame to confirm receipt of the data frame.

FIG. 1 is a block diagram of a Data Processing System where this invention may be practiced.

FIG. 2 is the format for a data frame for the modified HDLC protocol.

FIG. 3 is the format for a null frame in the modified HDLC protocol.

FIG. 4 is the data transparency encoding transformations for the modified HDLC protocol.

FIG. 5 is an address, control, and data byte for the modified HDLC protocol.

FIG. 6 is the format for an address byte for the modified HDLC protocol.

FIG. 7 is the format of the control byte showing the bit significance of the bits composing the byte.

with reference now to the figures, and in particular with reference to FIG. 1, there is depicted a pictorial representation of a data processing system 8 which may be utilized to implement the method of this invention. As may be seen, data processing system 8 may include a plurality of networks, such as Local Area Networks (LAN) 10 and 32, each of which preferably includes a plurality of individual computers 12 and 30, respectively. Of course, those skilled in the art will appreciate that a plurality of Interactive Work Stations (IWS) coupled to a host processor may be utilized for each such network.

As is common in such data processing systems, each individual computer may be coupled to a storage device 14 and/or a printer/output device 16. One or more such storage devices 14 may be utilized, in accordance with the method of the present invention, to store applications or other programs which may be periodically accessed by any user within data processing system 8.

Still referring to FIG. 1, it may be seen that data processing network 8 may also include multiple main frame computers, such as main frame computer 18, which may be preferably coupled to LAN 10 by means of communications link 22. The communications link 22 may further include modems (not shown) at each end of the link capable of simultaneous bidirectional data flows. Main frame computer 18 may also be coupled to a storage device 20 which may serve as remote storage for LAN 10. Similarly, LAN 10 may be coupled via communications link 24 through a subsystem control unit/communications controller 26 and communications link 34 to a gateway server 28. Gateway server 28 is preferably an individual computer or Interactive Work Station (IWS) which serves to link LAN 32 to LAN 10.

As discussed above with respect to LAN 32 and LAN 10, applications or other programs may be stored within storage device 20 and controlled by main frame computer 18. Of course, those skilled in the art will appreciate that main frame computer 18 may be located a great geographic distance from LAN 10 and similarly LAN 10 may be located a substantial distance from LAN 32.

As indicated, components of the above data processing system may be located substantial distances from each other. High-performance asynchronous (async) links may be required with specific sets of features, similar to those of this invention, to enable efficient transmissions over the communication links. The basic link concepts of this invention are derived from the High-Level Data Link Control (HDLC) protocol. The modifications meet the requirements of a high-performance async link using the limited data rates of low cost async modems. The link protocol of this invention requires very low overhead and allows application data to flow efficiently. On average, this invention adds only five bytes per frame to an application data packet. A packet, as used in this invention, refers to a sequence of binary digits, including data and controls signals, that is transmitted and switched as a composite whole. The data and control signals may also include error control information arranged in a specific format. The term packet is also synonymous with data frame.

At 1200 bits per second (bps), the target packet size is 60 bytes per frame, giving a framing overhead of approximately 8 percent. At 2400 bps the target packet size increase to 120 bytes, reducing the framing overhead to 4 percent. This is a full-duplex protocol, and was selected to be so, in order to lessen or eliminate the delay of the negative acknowledgement of bad frames. In typical synchronous protocols, half-duplex data flow can delay the negative acknowledgement of bad frames by up to six full frame times in a worst case situation. The full-duplex operation of this invention reduces the negative acknowledgement to a little over one full frame time in a worst case situation. Another positive effect of full-duplex operation, where bidirectional data flow is permitted, is a doubling of the bandwidth of the link by allowing data flow in both directions simultaneously.

Turning to FIG. 2, the format for a data frame shown. Two types of frames are defined by this invention--data frame and null frame--to provide for the transport of data packets across the link and for timely acknowledgement of their receipt. Application data will be encapsulated in a data frame consisting of an initial flag byte 40, an address byte 42, a control byte 44, one or more data bytes 46, two frame check sequence (CRC) bytes 48, and a terminal flag byte 50.

The address 42, control 44, data 46 and CRC 48 bytes are subject to a data-transparency transformation and will be expanded to include escape sequences if necessary before being queued to the physical transmitter. Upon receipt, these same bytes are processed to reverse the data-transparency transformation before being examined. Data transparency is achieved by the use of byte stuffing logic. Byte stuffing refers to the final operations necessary to physically transmit a frame. It refers to the addition of the begin/end recognitions indicators, as well as data transparency information. Byte stuffing occurs only after a transmit data packet has been framed, had the CRC calculated and inserted, and has been (optionally) encrypted. Byte stuffing logic first defines an "escape" character. Whenever the escape character flows over the link, it is always the first character of a two character escape sequence. The escape character is used to provide an alternate encoding mechanism for occurrence of the frame begin/end character in normal data and to encode actual occurrences of the escape character itself when they occur in the normal data. The frame start flag 40 is defined to be the eight-bit byte 0x7e. This value is guaranteed not to occur within the frame by the data transparency transformation which is applied to the frame image before it is queued to the physical transmission process. This value is also used for the frame end flag 50. No conflict occurs and by using the same value for the frame start 40 and end flag 50, it is possible to pack sequential frames end-to-end and allow the end flag 50 of the first frame to double as the start flag 40 of the next frame.

Limiting the number of eight-bit byte values which must be protected by the data transparency transformation minimizes the reduction of link capacity throughput caused by the expansion of conflicting data bytes into escape sequences. If we assume that the majority of the data flow is highly compressed, and that the compressed data will have an essentially "white" spectrum, the impact of data transparency expansion to avoid the two reserved bytes (the frame flag and the escape flag) should average less than one percent.

The escape flag is defined to be the eight-bit byte 0x7d. On the physical link, this value is guaranteed to be the first byte of a two-byte escape sequence in which the second byte will never be equal to either the escape flag value or the frame start flag 40 value. To achieve data transparency (unrestricted frame data content), escape sequences are substituted for any occurrences of the values 0x7e (the frame start flag 50) or 0x7d (the escape flag itself) in the address 42, control 44, data 46 and frame check sequence fields 48 of data frames. When one of these two special byte values is encountered, the transmitter substitutes an escape flag byte followed by the original byte value exclusive OR'ed with the byte value 0x20. This means transformations will occur as shown in FIG. 4. The receiver reverses the transformations to reconstruct the original byte values in the address 42, control 44, data 46 and frame check sequence 48 fields after the frame is received by the physical transport layer.

The integrity of each frame is statistically guaranteed by a 16-bit CRC 48 frame check sequence which is calculated and inserted in each frame on transmission and checked and deleted from each frame on receipt. Frames which fall the check sequence confirmation calculation are discarded just as through they had never been received. The frame check sequence consists of a two-byte cyclic redundancy check value calculated from the values of the address 42, control 44, and data 46 bytes before the data transparency transformation is performed. The CRC generator polynomial is defined by the 16 bit value 0x1021.

Turning to FIG. 3, the format for a null frame is shown. Data exchanged between two link stations takes the form of a series of data frames. These frames can be either the data frames as shown in FIG. 2, or the null frame of FIG. 3. A null frame carries no data, unlike the data frame which carries one or more bytes of application data. The link supervisory functions are handled by the exchange of null frames and application data frames.

Referring now to FIG. 5, details of the data frame's address byte 42, control byte 44 and data byte 46 will be described. Data frames consist of a sequence of 8-bit bytes, with the bytes transmitted and received using an eight-bit data word with one start bit, one stop bit, and no parity bit.

Every frame contains an address byte 42 which consists of eight individual receive toggle bits. Each receive toggle bit corresponds to one of the eight possible transmit sequence numbers from zero to seven. As shown in FIG. 6, the least significant bit (lsb) corresponds to transmit sequence number zero and the most significant bit (msb) corresponds to transmit sequence number seven.

Referring again to FIG. 5, whenever a frame is accepted by the receiver, the value of the receive toggle bits in the address byte 42 of the received frame are checked against the status of any unacknowledged transmitted frames. If the value of the appropriate receive toggle bit matches that expected for a particular transmitted frame, then that frame is considered to be acknowledged and the associated transmit sequence number becomes available for reuse.

This invention provides an async link protocol with a selective reject (SREJ) capability similar to that used in the HDLC protocol. Under HDLC, a re-transmission of the damages frames occur only when a single error is discovered in a sequence of consecutive frames. This invention improves on the selective reject capability by allowing up to six damaged frames in a sequence of up to seven consecutive frames to be simultaneously identified and selectively retransmitted as necessary. This invention also minimizes the impact of error transmissions on error recovery. A dynamic frame size adjustment algorithm is employed that decreases the frame size when multiple errors occur. This allows more frames to pass over the link between failures. A positive throughput is thus achieved despite the occurrence of multiple errors. If no errors occur for a period of time, the frame size is allowed to rise back to the optimal values.

The regular HDLC protocol station address supports several types of link topology including multi-drop and ring connections. Because of the operating environment--modems only--this modified HDLC protocol supports only point-to-point topology.

This modified HDLC protocol provides for sixteen independent logical sessions on a single physical link. This is logically supported by a dedicated four-bit field in the data frame header which is used to route packets to specific application queues. Although this function is not strictly part of the link layer, it is efficiently integrated into the link layer and eliminates the need for a separate low-level packet router task.

Referring again to FIG. 5, this invention uses a mechanism similar to that used by SDLC and HDLC to sequence frames. A three-bit transmit sequence number 80 is assigned to each frame. This invention however, never allows more than seven sequentially numbered frames to be unacknowledged at any one time. The transmitter can assure that data will not overrun the receiver and can automatically initiate retransmissions based on acknowledgements from the receiver.

Referring to FIG. 7, frames which carry data include a control byte 72 which contains three distinct fields. The first field is a four-bit logical channel index 76 which is used to route the data to any of sixteen distinct destination queues. The second is a three-bit transmit sequence number 80 field which is used to assure sequential delivery of data packets to target queues in the face of possible transmission errors which might cause physical reception of packets to occur out of the order in which they were originally supplied to the transmitter. The third field is a single-bit transmit toggle bit 78 value which works in conjunction with the transmit sequence number 80 to detect and discard multiple copies of data frames which may be received when an acknowledgement of the first copy is lost due to line noise.

The three-bit transmit sequence number 80 field is used to enforce serialization of data packets as they flow across the link. Under normal (low noise) conditions, packets will be received in the order in which they were scheduled for transmission. When errors occur which cause retransmissions to occur, it becomes necessary to transmit frames out of order.

As the transmitter schedules data packets for transmission, it assigns sequential sequence numbers in a radix 8 cycle from zero to seven and then back to zero. All eight sequence numbers start out in an available status. A total of no more than seven sequence numbers may be in use at any one time. For the transmitter to schedule a data packet for transmission, the next sequence number must be available and the total number of sequence numbers in use must be six or less.

Once a sequence number has been assigned, it can only be made available again if the receiver has confirmed reception of that sequence number and the sequence number immediately before it is already available. This has the effect of pacing data flow over the link during error recovery.

The transmit toggle 78 bit is inverted by the transmitter each time a sequence number is reused. The receiver confirms receipt of a packet each time its transmit toggle bit 78 is different than the toggle bit of the last packet which carried the same transmit sequence number.

If a packet is received with a transmit toggle bit which matches the last one received on a packet with the same transmit sequence number then the packet is discarded as a duplicate and is not confirmed.

The logical channel index 76 is logically treated as data by the link layer logic until a received packet is released by the link layer for delivery to the application layer. At this time, the value of the channel index 76 is use to index into a table of registered applications to determine the proper target queue for the incoming data.

This invention permits the release of error-free frames to their destinations immediately upon receipt. Those skilled in the art will know that such a technique is applicable in packet switched environments where higher levels of the interapplication protocol can re-sequence frames as required. This is the case, for example, in X.25 networks.

In summary, this invention discloses a new protocol and data packaging format which is optimized for noisy communications environments. A communications protocol utilizing full duplex communications channels is provided in which transmission errors are expected to occur. This invention does not require control messages to acknowledge or reject transmission. Instead, the control mechanism is incorporated into the normal data communications header. This enables the invention to handle multiple selective rejects of frames by increasing the amount of re-transmission control information conveyed to a sender. This invention provides for the re-transmission of only the frames in error by providing a packet format incorporating both data and control packets in a single control/data packet format. A subset of this new packet format is a "null" packet, which is defined for use in data-free feedback receipt confirmation signaling. It is expected that in actual implementation, the packets would be enveloped by frame flags and protected by a Cyclic Redundancy Code (CRC) or check sum. Depending on the type of communication channel, data transparency manipulations might also need to be performed. A receive toggle bit byte is provided consisting of an eight-bit value which encodes the toggle bit status for each of eight transmit sequence numbers. The receive toggle bit byte constitutes a null packet used for data packet receipt confirmation. The receive toggle bit byte is also the first byte of every data packet and is used here also for receipt confirmation. The second byte of a data packet is an eight bit byte consisting of a four-bit logical channel index, a single bit transmit toggle bit, and a three-bit transmit sequence number. The second byte is followed by zero or more eight-bit data bytes. Message confirmation is achieved by having a receiver, upon reception of a data packet, copy the bit value of the transmit toggle bit into the appropriate receive toggle bit based on the transmit sequence number. The receiver then relays the modified receive toggle bit byte back to the sender as part of a normal data packet or as a simple null packet. Thus, the invention is fully symmetrical, requiring no primary or secondary link station definition.

While the invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.