US 6,029,202 AGrant

Full Duplex Flow Control for Ethernet Networks

Issue Date:2000-02-22

•17 Claims

•10 Drawing Sheets

Abstract

CSMA/CD is used to implement flow control in a full-duplex Ethernet network in a lossless fashion. Uniquely identifiable flow control transmit on/off ("XON/XOFF") messages are transmitted, preferably during IPG, by a receiving station about to be congested to the transmitting station whose data output is to be controlled. The transmitting station physical layer receives and decodes these messages. If XOFF is recognized, the transmitting station continuously asserts CRS to its MAC layer at the MII, regardless of the prior CRS current state. CRS is continuously asserted until the receiving station transmits an XON flow control signal, indicating its ability to accept further data. During CRS assertion, the transmitting station defers transmission, e.g., is flow controlled. The MAC layer is slightly modified (but is still backward compatible with half-duplex networks) to provide separate transmit deferral receive data frame mechanisms using separate and independent input status signals, namely CRS and RX.sub.-- DV. CRS provides a carrierSense signal used for deferral within the MAC transmit process, and RX.sub.-- DV provides a receive.sub.-- carrierSense signal that frames data within the MAC receive process. As long as CRS remains asserted, the transmitting station defers transmission, thereby implementing flow control. In addition to slight MAC layer modification, the present invention slightly modified the physical layer, MII interface and reconciliation sublayer.

Metadata

Assignee

Sun Microsystems, Inc.

Inventors

Howard M. Frazier
Shimon Muller

Application Information

Application Number:US 9970428

Filing Date:1997-12-23

Priority Date:1995-11-06

Art Unit:276

Classifications

IPC:

G06F 1338G06F 1517

Field of Search:

709370232;233;235;237;234;250229;231;235;236

Patent Drawings (10 sheets)

Description

Field of the Invention

This invention relates to networks in general including Ethernet networks, and more specifically to implementing an Ethernet network having full duplex flow control

Background of the Invention

A network is a communications facility that permits a number of workstations, computers or other equipment (hereinafter collectively "computer(s)") to communicate with each other. Portions of a network involve hardware and software, for example, the computers or stations (which individually may comprise one or more central processing units, random access and persistent memory), the interface components, the cable or fiber optics used to connect them, as well as software that governs the access to and flow of information over the network. In networks in which data flow is 100 Mbits/sec. ("Mbps") or higher, the transmission medium is often fiber optics. In networks in which a slower data rate is acceptable, e.g., 10 Mbps, the transmission medium may be coaxial cable or, as is often the case for an Ethernet network, twisted wires.

In a network, network architecture defines protocols, message formats and other standards to which the computers and other equipment, and software must adhere. Most network architectures have adopted a model comprising functional layers in which a given layer is responsible for performing a specific set of functions, and for providing a specific set of services. Thus, the services provided by each layer and the interlayer interfaces can define a network architecture. Protocols define the services covered across a layer interface and the rules followed in the processing performed as a part of that service.

Several organizations have proposed models and standards that have been accepted within the networking community. The International Standards Organization ("ISO"), for example, has proposed a seven layer reference model for computer networking that is called the open systems interconnect ("OSI") architecture. Another set of standards has been promulgated by the Institute of Electrical and Electronics Engineers ("IEEE") set of proposed local area network ("LAN") standards known as IEEE Project 802. This model conforms to the seven-layer OSI model, but directly solely to the lowest two OSI layers, namely the physical layer and the data link layer.

FIG. 1 depicts a network according to the IEEE Project 802 modification to the ISO seven layer model, in which two computers 10, 10' are can communicate data to each other over a physical link medium 20, e.g., cable. Of course, in practice, a network may have many hundreds of computers rather than two.

The bottommost layer 30 in both the ISO and Project 802 model is a physical layer that is concerned with connections between two machines (e.g., computers 10, 10') to allow transmission of bit streams over a physical transmission medium (e.g., cable 20). Thus, physical layer 30 is concerned with types of cabling, cable plugs, connectors, and the like.

As will be described shortly, the present invention is directed to Ethernet networks adhering to the carrier sense multiple access with collision detection ("CSMA/CD") standard. In the 802 model for CSMA/CD, a reconciliation interface 40 defined by a Media Independent Interface ("MII") standard exists for the reconciliation sublayer 40 interface between physical layer 30 and a media access control ("MAC") sublayer 50B. The existing MII signal set provides independent four bit-wide paths for transmission and reception of data, and includes specific "hooks" for link level data flow control. (As used herein, flow control refers to inhibiting access to a network, or one or more links within the network.)

Interestingly, before adoption of MII the MAC standard defined a single carrier sense signal ("CRS") from the physical layer to the MAC that the MAC used to describe the state of the transmit and receive medium. This one CRS signal was used by the MAC transmit process to implement deferral of data transmission, and was by the MAC receive process to frame received data. With introduction of MII, this one CRS signal was decoupled into a CRS signal that again went to the MAC transmit process to implement deferral of data transmission, and into a receive data valid ("RX.sub.-- DV") signal that went to the MAC receive process. Thus, with MII, CRS is used solely by the MAC transmit process

Under MII, data and delimiters are synchronous to the corresponding clock, and two asynchronous media status signals are provided, namely carrier sense ("CRS"), and collision ("COL"). MII provides a two wire serial management interface for control and status gathering, namely management data clock ("MDC"), and management data input/output ("MDIO"). In the OSI seven-layer model, the layer above the physical layer is a data link layer that is responsible for error-free transmission of data frames between network nodes. A data link control protocol describes operation and interfaces of this layer, which must also shield higher layers in the model from concerns about the physical transmission medium.

But in the 802 model shown in FIG. 1, the data link layer is subdivided into MAC layer 50B and an overlying logical link control ("LLC") layer 50A. The media access control sublayer is concerned with access control methods to determine how to control the use of the physical transmission medium. The LLC sublayer 50A is responsible for medium-independent data link functions and allows the network layer 60 above to access LAN services independently of how the network is implemented. According to the 802 architecture, LLC sublayer 50A provides services to network 60 in the same fashion as would a conventional data link protocol in a wide area network.

The MAC sublayer 50B provides services to the overlying LLC sublayer 50A, and manages sharing of the transmission medium among the different stations on the network. A media access management function receives a frame from the data encapsulation function after the necessary control information has been added. Thereafter, media access management is responsible for ensuring physical transmission of the data. The data frame in an Ethernet full-duplex environment has a maximum size of 1,518 bytes.

Several 802 standards exist for MAC sublayer 50B, but only the carrier sense multiple access with collision detection ("CSMA/CD") standard is relevant to the present invention, more specifically the 802.3 standard. The existing 802.3 MAC standard presently contains several mechanisms for performing flow control in a half-duplex environment, including a Deference process, and WatchForCollission and BackOff procedures. CSMA/CD defines data encapsulation/decapsulation and media access management functions performed by MAC sublayer 50B it self, the data encoding/decoding function being performed by underlying physical layer 30.

Physical transmission of the data may be ensured using carrier sensing to defer transmission until the network is clear. In brief, a transmitting station (e.g., computer or user 10) listens or monitors the transmission medium (e.g., cable 20) before transmitting to determine whether another station (e.g., computer or user 10') is currently transmitting a message, e.g., to learn whether the medium is free. Using the services of the physical layer 30, the media access management determines whether the transmission medium (or carrier) is presently being used. If the medium is not being used, media access management passes the data frame to physical layer 30 for transmission. Even after transmission of the frame has begun, media access management continues to monitor the carrier. If the carrier is busy, media access management continues monitoring until no other stations are transmitting. Media access management then waits a specified random time to allow the network to clear and thereafter begins transmission.

But other station(s) having messages to send may all listen simultaneously, discern that the transmission medium appears quiet, and begin to transmit messages simultaneously. The result is a collision and garbled messages. If signal collision is detected, receiving stations ignore the garbled transmission, transmitting stations stop transmitting messages immediately and transmit a jamming signal over the medium. Following collision, each transmitting station will attempt to re-transmit after waiting for a random backoff-delay time period for the carrier to clear. Thus, a station transmitting must listen sufficiently long to ensure that collision has not occurred.

In FIG. 1, network layer 60 concerns the routing of data from one network node to another. It is the role of network layer 60 to route data between network nodes.

Transport layer 70 provides data transfer between two stations at an agreed upon level of quality once a connection is established between the stations. Transport layer 70 selects the particular class of service to be used, monitors transmission to ensure maintained service quality, and advises the stations (or users) if quality cannot be maintained.

Session layer 80 provides services that organize and synchronize a dialogue occurring between stations, and manages data exchange between stations. As such, session layer 80 controls when stations can send and receive data, based upon whether they can send and receive concurrently or alternately.

Presentation layer 90 ensures that information is presented to network users meaningfully, and may provide character code translation services, data conversion, data compression and expansion services.

Application layer 100 provides a mechanism for application processes to access system interconnection facilities for information exchange. The application layer provides services used to establish and terminate inter-user connections, and to monitor and manage the interconnected systems and the resources they employ.

Those skilled in the art will appreciate that the network shown in FIG. 1 may be half-duplex (or shared), or full-duplex. In half-duplex, there is a single shared communications path, and at any given time a station can transmit and monitor, or can receive, but can never simultaneously transmit and receive. In a half-duplex environment, MAC layer 50B uses a CSMA/CD algorithm, e.g., IEEE 802.3, such that a station desiring to use the network to transmit must first listen to learn whether the network is busy. If busy, the station desiring to transmit defers transmission but continues to monitor medium status while waiting for the network to become idle. If idle, the station can begin transmitting. While transmitting, the station continues to monitor the network to ensure that another station does not also begin to transmit at the same time, or transmit in an overlapping fashion, e.g., to ensure there is no collision. Similarly, when a station receives data, it is inhibited from simultaneously transmitting.

As such, the CSMA/CD algorithm provides a useful pacing mechanism that will inhibit or defer access of arrival new data into the half-duplex network until the network is adequately prepared. In addition, a so-called backpressure algorithm is also used in half-duplex environments to inhibit a station from sending data into the network by transmitting to that station. Receipt of the transmission causes the would be transmitting station to believe the network is busy, and to defer or inhibit transmission.

By contrast, in full-duplex, there are transmit and a receive communications paths, and one or more stations may transmit and receive simultaneously. The dual communications channel or path may in fact be multiple wires or cables, or a single wire or cable that simultaneously carries transmit and receive signals in both directions, perhaps using frequency division. Full-duplex networks can provide a higher data rate than half-duplex networks, often 100 Mbps, and are often preferred because of the more rapid communication rate.

Unfortunately, in going to a full-duplex environment, the IEEE 802.3 CSMA/CD MAC half-duplex mechanisms for collision avoidance and for transmission deferral must be abandoned. In a full-duplex environment, a transmitting station that in a half-duplex network would be subject to flow control is permitted to transmit at the same time a receiving station can transmit. The concept of "collisions" or avoiding collisions is meaningless in a full-duplex environment. Further, since a station desiring to transmit over a full-duplex network does not monitor the medium, half-duplex deferral and backpressure flow control procedures are also useless. Thus, the various half-duplex CSMA/CD data pacing mechanisms cannot be used in full-duplex to control or limit access to the network. However, a full-duplex network nonetheless needs some mechanism to inhibit a station from transmitting.

The need for flow control arises because a full-duplex network contains resources that are finite. Networks include devices such as switches, routers, bridges, etc. that connect two network segments together and move data between the segments. These devices typically include buffer memory to accommodate data rate adaptation between links carrying signals at different rates, e.g., perhaps 100 Mbps incoming data presented to a network link capable of handling only 10 Mbps, or to otherwise function when other resources are occupied. While such devices can temporarily store rapidly arriving data during a slower read-out, device memories can readily fill and congest unless some mechanism can halt or slow the entry of new data into the network, or at least the link including the congested device. (While half-duplex networks can also have links with different data rates, the problem is aggravated in full-duplex networks in which considerably more data flows, especially at 100 Mbps.)

Unfortunately, in the prior art there is no flow control mechanism for full-duplex Ethernet networks that avoid such data congestion and resultant data loss, without sacrificing data throughput. Flow control devices implemented in full-duplex networks at the transport layer are slow and inefficient. Transport layer mechanisms can detect a data congestion condition and slow transmission rates. Unfortunately, data throughput will be slower than the rate associated with the congested device using such flow control mechanisms. Thus, data output from a 10 Mbps device receiving data at 100 Mbps will be substantially less than the 10 Mbps rate associated with the device.

Thus, there is a need for a flow control mechanism for a full-duplex Ethernet network. Preferably such mechanism should provide CSMA/CD collision and deferral access to a full-duplex Ethernet network as a flow control method, while maintaining back compatibility with existing networks and standards. Such method should sense when a receiving station or device is about to be (or is) congested with data and should temporarily inhibit a transmitting station or device from transmitting new data until the present data can be handled.

To be compatible with existing networks and equipment and software, such flow control should conform to IEEE 802 standards. Further, to promote its wide use, such flow control mechanism should involve only minor modifications to such standards. Such mechanism should control flow such that the data rate from a slow device coupled to a fast device is the data rate capability of the slow device. Further, while flow control should include a "refresh" mode in which flow control signals are sent relatively often, an option for a time-out mechanism should be provided to deactivate flow control in the event a re-activate flow control signal is somehow not received by a transmitting station.

Further, there is a need in a full-duplex Ethernet network for a MAC layer whose transmitter process and receiver process are each responsive to separately and independently generated signals.

The present invention discloses such a flow control mechanism and such a MAC layer for a full-duplex Ethernet network.

Summary of the Present Invention

The present invention utilizes CSMA/CD to implement flow control in a full-duplex Ethernet network in a lossless fashion, without reliance upon a single signal to describe the state of both the transmitting and receiving process of a station that is to be flow controlled. Flow control transmit on/off ("XON/XOFF") messages are transmitted by a receiving station that is about to be congested, across the physical layer to the transmitting station whose data output is to be controlled. The XON/XOFF signals have a format distinguishable from normal data and preferably are transmitted in the inter-packet gap ("IPG") between data frames as frequently as every IPG. If the receiving station resources can receive at least a frame of additional data, XON is transmitted during the IPG, otherwise XOFF is transmitted. Alternatively, rather than send (e.g., "refresh") XON/XOFF after every single frame, a timer within the transmitting station may be started by receipt of an XOFF, and after expiration of a given time, transmission can re-start even if no XON was received (perhaps due to data corruption). The timer mechanism may be used with a per/IPG refresh control signal to ensure reestablishment of transmission should an XON control signal fail to be received by the flow controlled transmitting station.

At the physical layer of the transmitting station at the other end of the medium, the uniquely formatted flow control messages are received, decoded, and distinguished from each other and from data. If XOFF is recognized, the transmitting station asserts its carrier sense signal (CRS) to its MAC layer at the MII.

The present invention slightly modifies the MAC layer to provide separate and independent transmit deferral receive data frame mechanisms using separate and independent input status signals, namely CRS and RX.sub.-- DV. Within the slightly modified MAC layer, the CRS signal is used to provide a carrier sense signal that is used to perform a deferral function within the MAC transmit process. Further, when RX.sub.-- DV is asserted on the MII, MAC receive processing logic accepts and process data from the physical layer, and then passes the processed data to the logical link control layer. However, the modified MAC is fully back-compatible with half-duplex networks.

When threatened congestion is imminent, a receiving station sends XOFF to the physical layer of the transmitting station. Upon receipt of XOFF, the transmitting physical layer asserts CRS at the MII (or other interface), and continues to assert this status signal until XON is received. As long as CRS is asserted, the transmitting MAC will defer data transmission, and is thus flow controlled or inhibited from transmitting onto the network.

When receiving station resources permit accepting additional data, its physical layer emits XON over the medium to the transmitting station, preferably during the IPG during its own data transmission. (The fact that the receiving station may be congested does not affect its ability to transmit in the full-duplex environment.) The originally transmitting station receives and recognized XON, and de-asserts CRS, which permits its MAC to transit.

Preferably the receiving station sends an XOFF signal when it is within a frame or so of being congested. At the transmitting end, if an XOFF is received during frame transmission, XOFF does not take effect until that frame transmission is complete. Thus flow control is lossless, with .+-.one frame granularity.

Other features and advantages of the invention will appear from the following description in which the preferred embodiments have been set forth in detail, in conjunction with the accompanying drawings.

Brief Description of the Drawings

FIG. 1 depicts a network, according to the prior art;

FIG. 2 depicts a full-duplex switching network employing data flow control, according to the present invention;

FIG. 3A depicts the spatial relationship between signals associated with transmission of an XOFF message on MII', according to the present invention;

FIG. 3B depicts the spatial relationship between signals associated with transmission of an XON message on MII', according to the present invention;

FIG. 3C depicts the spatial relationship between signals associated with reception of an XOFF message on MII', according to the present invention;

FIG. 3D depicts the spatial relationship between signals associated with reception of an XON message on MII', according to the present invention;

FIG. 3E depicts the spatial relationship between signals associated with transmission of a RELEASE message on MII', according to the present invention;

FIG. 3F depicts the spatial window during which valid flow control messages may be sent, according to the present invention;

FIG. 3G depicts changes to MII' signal and CRS behavior during normal packet transmission, according to the present invention;

FIG. 3H depicts changes to MII' signal and CRS response behavior during XON/XOFF, according to the present invention;

FIG. 3I depicts changes to MII' signal and CRS response when XOFF is received during packet transmission, according to the present invention;

FIG. 3J depicts changes to MII' signal and CRS response in response to RELEASE, according to the present invention;

FIG. 4 depicts the role of the station management entity ("STA") within each DTE, according to the present invention;

FIG. 5 depicts modifications to and the mapping role of the reconciliation sublayer, according to the present invention;

FIG. 6 depicts the modified media access layer ("MAC") and its receipt of independent and separately generated CRS and RX.sub.-- DV signals, according to the present invention.

Detailed Description of the Preferred Embodiment

FIG. 2 depicts a full-duplex switched environment network employing data flow control to prevent packet congestion, according to the present invention. The network of FIG. 2 is in many ways similar to the prior art network of FIG. 1, except that minor changes have been made to the media access control sublayer, here 50B', to the reconciliation sublayer, here 40', to the interface standard, here MII', between the physical layer and the reconciliation sublayer 40', and to the physical layer, here 30'.

It will be appreciated that layers from and including the logical link control layer 50A and upward are unchanged, as are the computers 10, 10' and the interconnecting medium 20. Because only the MAC sublayer 50B', the reconciliation sublayer 40', the MII interface MII' and physical layer 30' are slightly changed, the network is fully IEEE 802 compatible, and is also compatible with the newly promulgated IEEE 802.3u standard. As a result, link level flow control is readily achieved without violating system capability with existing standards, including the 802.3 standard.

According to the present invention, a unit of data terminating equipment ("DTE"), e.g., computer 10, physical layer 30', interface MII', reconciliation sublayer 40', and MAC layer 50B sends an XON or XOFF message to flow control the DTE at the remote end of a link. With reference to FIG. 2, the second DTE might include computer 10' and associated relevant layers 30', 40' 50B' and interface MII'. As will be described, flow control messages XON/XOFF preferably are transmitted on the data path during the normally at least 96 bit times-wide inter-packet gap ("IPG") period, using a unique pattern of bits to distinguish XON and XOFF from any possible data. However, these flow control signals could instead be transmitted during frames of actual data, although reliability may be impacted as bit error might result in misidentification of a flow control signal. It is important that the transmitting station's physical layer distinguish incoming XON/XOFF (or other flow control messages, e.g., RELEASE) from data, and the flow control messages may use reserved codes, non-data codes, among other code patterns to be distinguished from normal data patterns.

Using the present invention, a DTE may transmit as long as it has not received an XOFF message from the DTE at the far end of the link. Upon receiving such an XOFF message, a DTE must first receive an XON message before it is allowed to transmit. In FIG. 2, the XON and XOFF messages are encoded and decoded by the physical layer 30'. Physical layer 30' responds to an XOFF by forcing an 802.3-compliant MAC layer 50B' to defer until receipt of an XON, or until a time-out occurs. The XON and XOFF messages are transported over the MII' interface via reserved encodings of the data path and delimiter signals. As described herein, the present invention uses the MAC 50B' deference process as a mechanism for lossless full-duplex flow control.

A DTE may initiate transmission of an XON or XOFF message via the following primitive:

RS.sub.-- FlowControl.request (request.sub.-- type)

request.sub.-- type: START,STOP

As used herein, the term "primitive" refers to a formal definition of a service interface provided by a model layer. For example, layers in communication with processes are defined by the services the layers provide to other layers and the individual specification of such services are primitives.

A DTE may receive an XON or XOFF message via the following primitive:

RS.sub.-- FlowControl.indicate(indicate.sub.-- type)

indicate.sub.-- type:START,STOP

A DTE may recover from a lost XON message via the following primitive:

RS.sub.-- FlowControl.sub.-- Release.request() XON and XOFF messages may be transmitted only during the IPG, or the first half of the preamble.

In the conventional MII signal set independent four bit-wide paths for transmission and reception of data are present, as are specific "hooks" for link level data flow control. The four transmission paths are transmit data or TXD<3:0>, transmit clock or TX.sub.-- CLK, transmit enable or TX.sub.-- EN, and transmit error or TX.sub.-- ER. TX.sub.-- CLK is generated by the physical layer and is used by the MAC as a data reference clock. TX.sub.-- CLK runs at 25% of the bit rate on the transmission media, e.g., 25 MHz for a 100 Mbps transmission rate and is continuously active. TX.sub.-- EN is MAC-provided, and frames or delimits data to be transmitted by the physical layer. TX.sub.-- EN is asserted into the active state ("1") at the start of a frame transmission, and remains active until de-assertion to logical "0" at end of frame transition. The transmit data TXD is actual frame data from the MAC to the physical layer. The TXD<3:0> nomenclature denotes that the transmit data is a four bit wide bundle. TX.sub.-- ER is a signal from the MAC to the physical layer advising that contents of the data bundle are not data code, e.g., perhaps corrupted data or XON, XOFF.

The four reception paths are receive data or RXD<3:0>, receive clock or RX.sub.-- CLK, receive data valid or RX.sub.-- DV, and receive error or RX.sub.-- ER. As noted, TX.sub.-- CLK is continuously active, whereas RX.sub.-- CLK is continuous active during data reception and an inter-packet gap3A ("IPG"). Within a protocol data unit ("PDU"), data transmission and reception is uninterrupted. As used herein, the notation TXD<3.0> or RXD<3,0> denotes that these signals are a four-bit wide bundle.

The production and handling of messages for the network of FIG. 2 will now be described. An XOFF message is produced by the reconciliation ("RS") sublayer 40' in response to the RS.sub.-- FlowContrrol.request(STOP)primitive. The XOFF message is conveyed on the MII' signals as follows:

TXD<3:0>=<0111>, TX.sub.-- EN de-asserted, TX.sub.-- ER asserted.

An XON message is produced by the RS 40' in response to the RS.sub.-- FlowControl.request(START) primitive. The XON message is conveyed on the MII' signals as follows:

TXD<3:0>=<1011>, TX.sub.-- EN de-asserted, TX.sub.-- ER asserted.

These XON and XOFF messages are conveyed in one TX.sub.-- CLK period.

FIG. 3A depicts the spatial relationship between TX.sub.-- CLK, TX.sub.-- EN, EXD<3:0>, and TX.sub.-- ER, with respect to transmission of an XOFF flow control signal from the MAC in the congested station. In FIG. 3A, a presumed unique code "0111" denotes XOFF, e.g., transmit data 3="0", transmit data 2, 1 and 0 are all "1". (The nomenclature "transmit data x" refers to a particular signal in the four-bit data bundle.) The receiving station's MAC provides TX.sub.-- EN, which is "0" as the waveforms shown are occurring during the IPG, and also provides TX.sub.-- ER from the reconciliation sublayer. The receiving station's MAC will also assert TX.sub.-- ER for one cycle of TX.sub.-- CLK while the "0111" pattern is placed on the data bundle to indicate this is a non-data signal. If TX.sub.-- ER is "1", the occurrence of 0111 in data TXD will look like data, but when TX.sub.-- EN="0" and TX.sub.-- ER="1", the unique pattern (here 0111) denotes XOFF. The "XX" notation for the data TXD indicates "don't care", since normal data has no meaning during the IPG time.

FIG. 3B depicts the spatial relationship between the same signals for transmission of an XON flow control signal (here denoted as "1011") from the no-longer congested receiving station MAC. TX.sub.-- EN is "0" because, as noted, XON preferably is transmitted between frames in the IPG. The no longer congested station's MAC provides the 1011 XON signal, which when TX.sub.-- EN="0" and EX.sub.-- ER="1" denotes XON, and not data.

FIG. 3C and FIG. 3D similarly depict reception of an XOFF and an XON message, by the transmitting station whose data output has threatened to congest the receiving station. In FIGS. 3C and 3D, RX.sub.-- DV ="0" because the various transmitting station's physical layer-generated waveforms shown occur within the IPG, and RX.sub.-- DV normally remains de-asserted during the IPG. In FIG. 3C, when a flow control message is received during the IPG, RX.sub.-- DV remains "0", the received data bundle assumes the value, here, "0111" denoting XOFF, and RX.sub.-- ER is asserted to a "1" state for one cycle of RX.sub.-- CLK. Similarly, with respect to FIG. 3D, to receive an XON message, RX.sub.-- DV remains de-asserted, the received data bundle assumes the state, here, "1011" denoting XON, and the normally deactivated RX.sub.-- ER is raised to "1" for one cycle of RX.sub.-- CLK.

The physical layer 30' sends an XOFF message to the RS 40' when the physical layer received an XOFF from the remote end of the link. The XOFF message is conveyed on the MII' signals as follows:

RXD<3:0>=<0111>, RX.sub.-- DV de-asserted, RX.sub.-- ER asserted The RS generates FS.sub.-- FlowControl.indicate(STOP) when it receives an XOFF message.

The physical layer sends an XON message to the RS when the physical layer receives an XON from the remote end of the link. The XON message is conveyed on the MII' signals as follows:

RXD<3:0>=<1011>, RX.sub.-- DV de-asserted, RX.sub.-- ER asserted The RS generates RS.sub.-- FlowControl.indicate(START) when it receives an XON message. The XON and XOFF messages are conveyed in one RX.sub.-- CLK period.

In a preferred "refresh mode" embodiment, XON or XOR is transmitted by a receiving station preferably during each IPG, although these flow control signals could instead be transmitted with a repetition less than once per IPG. In an alternative "time-out" embodiment, a timer, shown in FIG. 2 as 110, associated with the MAC layer times an interval .DELTA.T commencing with receipt of an XOFF flow control signal. If after interval .DELTA.T an XON signal has not been received (perhaps due to corruption or other error), the transmitting station's MAC issues a RELEASE signal ("1100" in FIG. 3E) that forces the underlying physical layer to de-assert the CRS signal. The MAC can then transmit, precisely as if an XON signal had been received. Although implementing timer 110 involves some additional cost and hardware, its use can make available many of the IPG periods for use other than for conveying flow control messages. On the other hand, transmitting flow control messages during every IPG is a more robust procedure, but dedicates at least a portion of the IPG period to flow control message transmission.

The RELEASE message is generated by the transmitting station's reconciliation sublayer 40' in response to the RS.sub.-- FlowControl.sub.-- Release.request() primitive. The RELEASE message is conveyed on the MII' signals as follows:

TXD<3:0>=<1100>, TX.sub.-- EN de-asserted, TX.sub.-- ER asserted.

This RELEASE message is conveyed in one TX.sub.-- CLK period.

FIG. 3E depicts the spatial relationship between TX.sub.-- CLK, TX.sub.-- EN, TXD<3:0>, and TX.sub.-- ER with respect to coding of a RELEASE ("1100") message on MII'. TX.sub.-- EN="0" as what is shown is still the IPG period. Assertion of TX.sub.-- ER from "1" to "1" during one TX.sub.-- CLK cycle while "1100" is present on the data bundle results in generation of a RELEASE signal from the MAC to the physical layer. As noted, the physical layer de-asserts CRS, which enables MAC to transmit.

The RS 40' must ensure that flow control messages are only sent during the IPG or during the first nibble of the preamble. Messages may be sent when the following conditions are met:

TX.sub.-- EN.sub.[t] =0 OR TX.sub.-- EN.sub.[t-1] =0.

FIG. 3F depicts this spatial relationship transporting flow control messages across MII' when emitting XON, XOFF, or RELEASE. Thus, FIG. 3E is relevant to both the refresh mode and time-out mode of generating flow control signals. In FIG. 3F, although XON is shown on TXD, in fact, XOFF or RELEASE could also have been shown. Two time windows for possible emission of a flow control signal are shown in FIG. 3F, denoted "earliest" and "latest".

In FIG. 3F, the earliest emission of a flow control signal is in the nibble time or clock period following the last nibble of a data frame. (A "nibble" is a four-bit unit of data that requires 25% of the nominal bit rate to be sent across the MII'.) The latest opportunity for sending a flow control message is overlapped with the first nibble of a data packet. This result occurs because the flow control messages are presumed emitted by an entity that is logically separate from the normal MAC transmit processes.

The present invention does not require changes to the MAC transmitting process to emit flow control messages, but instead assumes a parallel process logically separate from the MAC transmitting process. While this assumption promotes protocol compatibility, there is no correlation between sequencing of the MAC transmission processes and the state of the process that emits the flow control messages. Stated differently, the two processes neither know nor can predict each other's actions for the next TX.sub.-- CLK cycle. Thus, the MAC could start transmitting a data packet when the process wishing to signal flow control begins to send XON, XOFF, or RELEASE. To avoid interference, the flow control generating process can monitor the state of TX.sub.-- EN and if TX.sub.-- EN is "0", or if it was de-asserted to "0" on the previous TX.sub.-- CLK cycle, then a flow control message may be sent. But if TX.sub.-- EN is asserted and was asserted on the previous TX.sub.-- CLK cycle, then no flow control message can be sent. Thus, the process sending flow control messages must monitor TX.sub.-- EN and may only send flow control messages when TX.sub.-- EN is de-asserted, or when it was de-asserted during the previous TX.sub.-- CLK signal.

FIG. 3G depicts the response of the CRS and collision ("COL") signals in response to TX.sub.-- EN during normal packet transmission.

TX.sub.-- EN is generated by the MAC to the physical layer at the MII to indicate that a data frame is being transmitted, and the CRS and COL are signals from the physical layer to the MAC. CRS is asserted by the physical layer in response to the transmission. CRS remains asserted during transmission, and de-asserts after TX.sub.-- EN de-asserts. The MAC defer process, a portion of a state machine associated with the MAC layer, begins time the IPG with the de-assertion of CRS. COL remains inactive or "0" at all times for a full-duplex environment.

By contrast, if FIG. 3G was associated with a half-duplex link, no simultaneous reception would be occurring. In half-duplex, the physical layer 30' asserts CRS during packet transmission or reception, whereas in full-duplex mode, the physical layer asserts CRS during packet transmission but not during packet reception. In full-duplex mode, the physical layer asserts CRS when an XOFF message is received from the remote end of the link, and continuously asserts CRS until an XON message is received from the remote end of the link, or until a RELEASE message is received from the local reconciliation layer 40'.

FIG. 3H depicts response of the CRS and collision ("COL") signals in response to TX.sub.-- EN in the presence of XOFF and XON signals. TX.sub.-- EN is de-asserted at the normal end of a data frame transmission, whereupon CRS de-asserts in response. But in FIG. 3H, some time after CRS ="0", the physically layer at the transmitting station receives an XOFF flow control signal from the congested receiving station at an end of the link. XOFF is decoded and in response CRS is re-asserted at the MII by the transmitting station, as shown in FIG. 3H. CRS stays active for an indeterminate period of time, during which time TX.sub.-- EN is low because the transmitting MAC is deferring and will not transmit when CRS remains asserted. Later in FIG. 3H, an XON signal is received by the transmitting physical layer, which de-asserts CRS, whereupon the MAC may start transmitting. But the MAC will first time the IPG (96 bit times), and then begin transmitting, and assert TX.sub.-- EN (if it has data to assert). The data to be sent is put onto the medium and the physical layer also asserts CRS.

FIG. 3I depicts the effect of receipt of an XOFF while a transmission is occurring, and demonstrates packet boundary granularity. MAC is transmitting a frame as shown by TX.sub.-- EN="1", and CRS="1". Sometime into the frame transmission, an XOFF signal is received. The MAC continues to transmit, terminating the packet normally. But upon conclusion of frame transmission, instead of de-asserting CRS (as in FIG. 3G), CRS remains asserted because XOFF was received, and remains asserted until XON is received. After XON is received, after an IPG, CRS is re-asserted and the MAC can begin transmitting again.

FIG. 3J depicts the response of the physical layer to a RELEASE message, and may be used if an XON message is somehow corrupted. The transmitting physical layer receives an XOFF message during transmission, which activates CRS for a relatively lengthy time period and inhibits the MAC from transmitting (e.g., deferral). The MAC continues to defer, no XON signal is received (perhaps due to signal corruption). After a period of time .DELTA.T, a RELEASE signal is received, which forces the physical layer to release or de-assert CRS, allowing the MAC to again transmit. The RELEASE message is encoded on the transmit path of the MII, forcing the physical to release CRS.

It is important to appreciate that while XOFF was received from a remote (congested) station perhaps at the end of a link, the RELEASE signal is received from the near end of the link, namely from a station management entity ("STA") 150 in the transmitting station (see FIG. 4).

Turning now to FIG. 4, a first piece of data terminating equipment or DTE, e.g., computer 10, physical layer 30', interface MII', reconciliation sublayer 40', and MAC layer 50B is shown coupled to a second DTE, e.g., computer 10' and corresponding layers and sublayers. Each DTE includes a station management entity ("STA") 150, which essentially is a set of management or control functions within the station or computer. The STA becomes aware when station resources (e.g., memory) are in danger of congesting. For example, as a design choice it is useful to signal congestion and cause an XOFF control flow signal to issue when resources have no more than about one frame of data storage available, although other thresholds could instead be set.

In FIG. 4, the MII' interface, physical layer signal ("PLS") service primitives, reconciliation sublayer service primitives, and flow control messages are generally depicted, as is the physical link, e.g., cable 20, between the DTEs. The PLS service primitives are provided by the physical layer 30' to the media access control layer 50B. As may be appreciated from FIG. 5, the reconciliation sublayer 40' functions to map the concrete signals defined by the MII' to the less concrete PLS service primitives.

For example, if a DTE STA 150 detects imminent resource congestion and wishes to emit a flow control signal (e.g., XOFF), the STA uses a RS STOP service primitive. The STA sends this primitive to RS 40', which encodes the primitive in MII'. The MII' then causes the associated physical layer 30' to emit the appropriate control flow signal, here XOFF.

It is important to appreciate from FIG. 4 that STA 150 is a separate and independent process from MAC 50B', and that no correlation exists between them. As a result, it is entirely possible that limited resources in computer 10' may result in its STA 150 determining that imminent congestion exists and that an XOFF signal must be transmitted, even though computer 10' may be ready to simultaneously transmit data. Thus, simultaneously a data packet from MAC 50B' and a flow control primitive from STA 150 could issue, both signal sets arriving as RS 40' simultaneously. Understandably, RS' will try to encode both signals simultaneously, but the result is that the physical layer 30' will transmit a garbled signal as it cannot discern the true data from the flow control signal. Thus, as was described with respect to FIG. 3F, the present invention creates a rule to avoid such overlap by specifying a window within which flow control primitives from MII' be encoded.

FIG. 5 will now be described with respect to changes to the MII' interface and the behavior of its associated signals. In FIG. 5 service primitives coupled to the reconciliation sublayer 40' include four new service functions, defined according to the present invention. Shown in bold type in FIG. 5, the new services are:

RS.sub.-- FLOW.sub.-- CONTROL.sub.-- RELEASE.request (input)

RS.sub.-- FLOW.sub.-- CONTROL.request (input)

PLS.sub.-- RX.sub.-- CARRIER.indicate (output)

RS.sub.-- FLOW.sub.-- CONTROL.indicate (output)

These new reconciliation service layer primitives contain processes, only the most relevant of which are depicted in FIG. 5. The various service primitives shown on the left side of FIG. 5 are given in standard primitives language. The MII' signals on the right side of FIG. 5 reflect the MII' interface specification. The MII' specification is merely the definition of the behavior of a set of signals at the interface between physical layer 30' and reconciliation sublayer 40', and the MII' as such is not a functional entity. Thus, as modified according to the present invention, MII' defines an additional set of signal behaviors, e.g., the encoding of a transmit data bundle (TXD>3:0>) when flow control signals such as XON/XOFF are required, or the behavior of the physical layer 30' in terms of its handling of a CRS signal, and so forth. What is modified is the behavioral description of such signals, and how such signals may be used to achieve flow control in a full-duplex Ethernet environment, without data loss.

FIG. 6 depicts in further detail the duality aspect of the MAC 50B' layer according to the present invention. As noted, in half-duplex MAC layers, a single CRS input signal from the physical layer controlled the MAC layer transmitter process deferral and also the MAC layer receive process data framing. In stark contrast, a modified MAC layer 50B' according to the present invention separates these two processes and uses independent and separate input signals to control the deferral and data framing mechanisms.

As shown in FIG. 6, within MAC 50B', the transmitter process 130 includes a deference mechanism 150 that response to a carrier Sensesense signal, and the receiver process 140 includes a BirReceiver mechanism 160 that responds to a new variable, defined according to the present invention, receive.sub.-- carrierSense signal. As shown, the receive.sub.-- carrierSense variable may be derived directly from the MII' signal RX.sub.-- DV, and is used to indicate incoming bits. (By contrast, in the prior art, a variable carrierSense was derived directly from the MII signal CRS and was used to indicate transmission deferral.)

In the preferred embodiment, separate CRS and RX.sub.-- DV signals at MII' are provided to RS 40', which outputs among the various service primitives shown in FIG. 5, the PLS.sub.-- carrier.indicate and PLS.sub.-- receive.sub.-- carrier.indicate primitives. The MAC 50B' receives these service primitives from which are provided as carrierSense and receive.sub.-- carrierSense to the deference and BitReceiver mechanisms respectively. This separation of MAC carrierSense variables into carrierSense (used by deference 150) and receive.sub.-- carrierSense (used by BitReceiver 160) is in contrast to the prior art.

For purposes of the present invention, it suffices that separate and independently generated signals give rise to carrierSense and receive.sub.-- carrierSense within the modified MAC layer 50B'. Stated differently, it need not be required that CRS and RX.sub.-- DV give rises to these signals. Regardless of how the separate and independently generated signals are created, in response the modified MAC layer 50B' performs separate deferral and data framing procedures similarly to what is accomplished for a half-duplex network using a MAC layer according to the prior art. Further, a modified MAC layer 50B' is fully backwards compatible with prior art networks that use the same CRS and RX.sub.-- DV signals. The present invention slightly and semantically modified the 802.3 MAC standard as follows:

The interface to the physical layer is as follows, wherein bold type indicates newly added features, e.g., receive.sub.-- carrierSense is a newly defined variable:

State Variable Initialization according to the present invention may be implemented as follows:

Frame Reception preferably is implemented as follows:

The present invention makes slight modification to the reconciliation sublayer as follows:

As shown in FIG. 6, PLS.sub.-- RX.sub.-- CARRIER.indicate(CARRIER.sub.-- STATUS) will map the MII' signal RX.sub.-- DV to the MAC variable receive.sub.-- carrierSense, and PLS.sub.-- CARRIER.indicate (CARRIER.sub.-- STATUS) will map the MII' signal CRS to the MAC variable carrierSense.

RS.sub.-- FLOW.sub.-- CONTROL.request (REQUEST.sub.-- TYPE) is generated by the STA to the reconciliation sublayer, and the REQUEST.sub.-- TYPE parameter can take one of two values: STOP and START. Upon receipt of this primitive, the reconciliation sublayer maps the STOP and START values to messages on the MII' as follows:

STOP->XOFF

START->XON

In the present invention, RS.sub.-- FLOW.sub.-- CONTROL.indicate (INDICATE.sub.-- TYPE) is generated by the reconciliation sublayer to the STA. The INDICATE.sub.-- TYPE parameter can take one of two values:

STOP and START

The reconciliation sublayer generates this primitive by decoding messages from MII and mapping them as follows:

XOFF->STOP

XON->START

The RS.sub.-- FLOW.sub.-- CONTROL.sub.-- RELEASE.request() primitive is generated by the STA to the reconciliation sublayer and takes no parameters. Upon receipt of this primitive, the reconciliation sublayer generates a RELEASE message on the MII'.

To recapitulate, it will be appreciated from all of the foregoing, that the present invention implements flow control in a full-duplex Ethernet environment, while preserving backward compatibility with existing environments and hardware. Flow control is provided using a slightly modified MII' signal set, with minor changes to the MAC layer 50B', with minor additions to the reconciliation sublayer 40' and with minor changes to the physical layer 30'.

The present invention provides deferral and data framing preferably using two independently generated signals. The resultant flow control is lossless and permits a reduced data rate equal to the date rate of the slower resource whose imminent congestion gives rise to the flow control. In short, link level flow control for full-duplex Ethernet networks may be implemented in a simple and cost-effective fashion, while maintaining backward compatibility.

Modifications and variations may be made to the disclosed embodiments without departing from the subject and spirit of the invention as defined by the following claims.

Claims

What is claimed is:

1. A method for providing flow control in a network that includes a first communicating device having a first media access control layer and a first physical layer, and a second communicating device, the first and second communicating devices being connected by a communication medium, the method comprising the following steps: (a) receiving data at the first physical layer from the first media access layer and transmitting the received data over the communication medium to the second communicating device; (b) continuously transmitting data received at the first physical layer from the first media access control layer to the second communicating device until the second communicating device transmits a first flow control signal to the first physical layer; and (c) generating a delay signal at the first physical layer in response to said first flow control signal and providing the delay signal to the first media access control layer, wherein the first media access control layer ceases providing the first physical layer with data for transmission to the second communicating device in response to the delay signal, thereby ceasing transmission of data between the first communicating device and second communicating device.

2. The method of claim 1, wherein the second communicating device transmits a second flow control signal to the first physical layer to resume transmission of data between the first and second communicating devices.

3. The method of claim 2, wherein the first physical layer generates a resume transmission signal in response to receipt of the second flow control signal, the first media access control layer being responsive to said resume transmission signal to resume providing data to said first physical layer for transmission to said second communicating device over said communication medium.

4. The method of claim 1, wherein the delay signal comprises a carrier sense signal.

5. The method of claim 1, wherein the network comprises an ethernet network.

6. The method of claim 5, wherein the second communicating device includes a second media access control layer and second physical layer.

7. The method of claim 5, wherein the first communicating device further includes an interface between the physical layer and the media access control layer.

8. The method of claim 7, wherein the interface comprises a media independent interface.

9. The method of claim 1, wherein the communication medium comprises a cable.

10. A system for performing flow control in a full duplex network that includes a first communicating device and a second communicating device interconnected through a communication medium, comprising: a first physical layer in the first communicating device for providing a full duplex connection of the first communicating device to the communication medium; a first media access control layer in the first communicating device for providing a flow of data to the first physical layer for transmission to the second communicating device, the first media access control layer being operable in a half duplex mode; and a flow control mechanism for controlling the transmission of data between the first and second communicating devices, wherein the first physical layer responds to the flow control mechanism to control the flow of data between the first media access control layer and the first physical layer utilizing the half duplex mode of operation of the first media access control layer.

11. The system of claim 10, wherein the flow control mechanism comprises at least one of a first flow control message and a second flow control message, the first physical layer preventing the flow of data from the first media access control layer to the first physical layer in response to a first flow control message from the second communicating device.

12. The system of claim 11, wherein the first physical layer prevents the flow of data by providing a delay signal to the first media access control layer in response to the first flow control message.

13. The system of claim 12, wherein the delay signal comprises a carrier sense signal.

14. The system of claim 11, wherein the first physical layer allows the first media access control layer to continue providing the flow of data to the first physical layer in response to a second flow control message from the second communicating device.

15. The system of claim 10, wherein the communication medium comprises a cable.

16. The system of claim 10, wherein the second communicating device includes a second physical layer and a second media access control layer, the second media access control layer being operable in a half duplex mode for providing a second flow of data to the second physical layer for transmission to the first communicating device.

17. The system of claim 16, wherein the flow control mechanism comprises at least one of a transmit-off message and a transmit-on message, the second physical layer delaying a flow of data from the second media access control layer to the second physical layer utilizing the half duplex mode of operation of the second media access control layer in response to a transmit-off message.

Patent Citations (16)

Patent	Date	Inventor
US4926415	1990-05-01	Tawara et al.
US5007051	1991-04-01	Dolkas et al.
US5121382	1992-06-01	Yang et al.
US5313582	1994-05-01	Hendel et al.
US5404353	1995-04-01	Ben-Michael et al.
US5418784	1995-05-01	Ramakrishnan et al.
US5526355	1996-06-01	Yang et al.
US5541957	1996-07-01	Lau
US5568476	1996-10-01	Sherer et al.
US5577069	1996-11-01	Lau et al.
US5600798	1997-02-01	Cherukuri et al.
US5745790	1998-04-01	Oskouy
US5784559	1998-07-01	Frazier et al.
EPX310360	1989-04-01
EPX333225	1989-09-01
EPX404337	1990-12-01

Non-Patent Literature (3)

"IEEE Standards for Local Area Networks: Logical Link Control"; The Electr. Engs. Inc.; pp. 21-76; Std 802.2, Nov. 1988.
Somer, G.; "Ethernet Transceiver Offers Upgrade from Existing Networks"; Electronic Engineering; vol. 67, No. 820; pp. 25-30, Apr. 1995.
Tanenbaum, A.S.; "Computer Networks"; Prentice-Hall Intl. Inc.; Sec. 1.4:14-19; Sec. 3.4:141-147; and Sec. 4.4:223-225, Mar. 1993.

Abstract

Description