WO1997011540A1

WO1997011540A1 - Method and apparatus for controlling flow of incoming data packets by target node on an ethernet network

Info

Publication number: WO1997011540A1
Application number: PCT/US1996/015180
Authority: WO
Inventors: Wang Yun-Che; David Wong
Original assignee: Cypress Semiconductor Corporation
Priority date: 1995-09-22
Filing date: 1996-09-23
Publication date: 1997-03-27

Abstract

In a CSMA/CD Ethernet network, an Ethernet interface chip (28) of a target node (15, 12) is equipped with a jam signal (38) for enabling the target node (15, 12) to control the flow of incoming data packets from the channel (16). The jam control signal (38) provides a mechanism for the target node (15, 12) to force collisions on the channel (16), forcing the source node (14) to retransmit the 'rejected' data packets at a later time, as determined by a predefined 'backoff' algorithm. When a jam signal (38) is activated, and the target node (15, 12) begins receiving incoming data packets from the channel (16), the target node (15, 12) refuses acceptance of all incoming data packets by immediately and automatically transmitting a dummy packet across the channel (16), intentionally forcing a collision, and causing the source node (14) to 'backoff' and retransmit at a later time. The dummy data packets are transmitted by the target node (15, 12) at a physical layer of the Ethernet protocol, bypassing standard Ethernet transmission protocol of submitting the dummy data packet to a media access controller (26).

Description

METHOD AND APPARATUS FOE COHTROLLIHG FLOW OF INCOMING DATA PACKETS BY TARGET HOPE OW AH ETHERNET NETWORK

BACKGROUND OF THE INVENTION

1. Field Of The Invention This invention relates generally to local area networks (LANs) of the Ethernet type and, more particularly, to a technique for a given node attached to an Ethernet LAN to delay the flov of incoming data by forcing collisions on the network channel.

2. B ckground Of Th* γt.

Ethernet is a commonly used name for a LAN that uses a network access protocol referred to as Carrier Sense Multiple Access with Collision Detection (CSMA/CD). The CSMA/CD protocol is defined in ANSI/IEEE std 802.3, published by the Institute of Electrical and Electronics Engineers, Inc., 345 East 45th Street, New York, N.Y. 10017.

Under the CSMA/CD rules for access to a network bus or cable, which will be referred to as the channel, any node or station wishing to transmit data packets must first "listen" to make sure that the channel is clear before beginning to transmit. All nodes on the network have equal priority of access and may begin transmitting as soon as the channel is clear and a required inter-packet delay of 9.6 microseconds has elapsed. A "collision" occurs when a first node and a second node both attempt to commence transmission of data packets on a clear channel at substantially the same time. When such a collision occurs, the first node continues transmitting for a short time to make sure that all nodes wishing to transmit will detect the collision. Every other transmitting node detecting the collision also continues to transmit for a short time. Then each node that has detected a collision terminates transmission of its data packet. The nodes involved in the collision wait for the required interpacket delay of 9.6 microseconds and then determine a respective random backoff time, as calculated by the IEEE 802.3 standard procedure (i.e., the so-called "truncated binary exponential backoff" algorithm), before attempting to retransmit the same data packet. According to standard Ethernet protocol, this sequence is repeated until the data packet that experienced the collision is either successfully transmitted or the data packet experiences a maximum of 16 retransmission attempts, at which time the data packet is "lost" and must be recovered at a higher level above the Ethernet protocol.

The standard Ethernet protocol provides that a first node (the "source") is able to transmit data packets to a second node (the "target"), provided that no other node on the network is currently transmitting data packets on the channel. However, the standard Ethernet protocol does not provide a mechanism for the target node to "throttle" the flow of incoming data, i.e. to force the source node to delay the transmission of data packets until it is desirable for the target node to accept the data packets. If a transmitted data packet is addressed to a particular target node, that target node must accept that data packet. If the target node is not capable of accepting the transmitted data packet, the data packet is "lost" and the source node is not notified of the lost packet and it must be recovered by the system at a higher layer above the Ethernet protocol, typically the network layer. Furthermore, standard Ethernet protocol prohibits a node from attempting to transmit data packets across the channel while it is receiving data packets from the channel.

Although a lost data packet will eventually be recovered at a higher layer above the Ethernet protocol and retransmitted by the source node, such a "recovery" process is slow and inefficient. First, significant time is expended during the "recovery" process at the network layer in determining whether one or more data packets have been "lost." Second, the "recovery" process at the network layer depends upon the target node signaling to the source node that the target node did not receive as many data packets from the channel as the source node transmitted across the channel. When the source node receives this signal, it must retransmit all data packets included in the original data packet "group", of which the "lost" data packet was a member, even though only one of the several data packets may have been "lost." Thus, time is wasted retransmitting data packets that were successfully transmitted the first time.

By way of example, a network "hub" is a node on an Ethernet LAN which monitors and processes the transmission of data "traffic" on the channel. All data packets transmitted across the network channel must pass through the hub node. Thus, the hub node may become particularly busy receiving and transmitting data packets. Although the hub node typically incorporates one or more buffers to temporarily store incoming data packets, it is possible for the buffer to become temporarily full and unable to accept additional incoming data packets. Under standard Ethernet protocol, once the hub node buffer is full, subsequent incoming data packets are not received by the hub node and are thus "lost." Once the source node transmits a data packet, it has no way of retrieving the data packet, even if it is rejected by the hub node, except through the recovery process at a higher layer above the Ethernet protocol. Thus, standard Ethernet protocol requires the source node to eventually retransmit the entire data packet group which includes the "lost" data packet.

Accordingly, there is a need to provide a mechanism for a target node to "throttle" incoming data packets that minimizes or eliminates one or more problems in the art as described above.

SUMMARY OF THE INVENTION

It is the objective of this invention to provide a method for a target node in a CSMA/CD Ethernet network to control the flow of incoming data packets from the channel. A preferred embodiment of the present invention accomplishes this objective by determining when it is desirable for the target node to limit the flow of incoming data packets and activating a jam signal in response thereto. The method further determines when incoming data is pending on the channel and activates a receive status signal in response thereto. When the jam signal and the receive status signal are both activated contemporaneously, the target node automatically sends a dummy data packet across the channel, thus forcing a collision on the channel and causing the source node to "backoff" from transmission. The dummy data packet is transmitted by the target node at the physical layer of the Ethernet protocol, bypassing the standard Ethernet transmission method of submitting data packets to a media access controller for transmission. It is a further objective of this invention to provide an apparatus for accomplishing this method of controlling the flow of incoming data packets from the channel.

These and other features and objects of this invention will become apparent to one skilled in the art from the following detailed description and the accompanying drawings illustrating features of this invention by way of example.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. IA is a simplified block diagram view of a four-node Ethernet network, illustrating four nodes of the same type connected by a shared medium cable or channel, wherein the - shared medium is a coaxial cable.

FIG. IB is a simplified block diagram view of a four-node Ethernet network, functionally equivalent to that shown in FIG. IA, illustrating four nodes of the same type connected to each other by shared mediums traveling through a hub node, wherein the shared mediums are pairs of twisted-pair wires.

FIG. 2 is a simplified functional block diagram view of one of the nodes of FIG. 1.

FIG. 3 is a simplified functional block diagram view of the Ethernet Interface Circuit shown in FIG. 2.

FIG. 4 is a simplified flow chart illustrating the steps performed by an embodiment of the present invention. FIG. 5 is a simplified flow chart illustrating, in greater detail , determining step 62 of FIG. 4.

The present invention relates to a target node's ability to throttle incoming data packets at the physical layer of the Ethernet protocol by forcing a collision on the channel. FIG. IA and FIG. IB both illustrate a simple Ethernet network having four identical nodes, node 10, node 12, node 14, and node 15, and channel 16. Fig. IB further illustrates a hub node 17. FIG. 2 illustrates in more detail node 15 which is of the same type as nodes 10, 12, and 14. FIG. 2 illustrates transmit buffer 18, receive buffer 20, memory 22, a control means or controller, such as central processing unit (CPU) 24, media access controller (MAC) 26, Ethernet interface circuit 28, transmit status signal 32, receive status signal 34, jam control signal 38, and data buses 30, 31, 36, and 37. As shown in FIG. 3, Ethernet interface circuit 28 includes data transmitter 40, means, such as jam transmitter 42 for transmitting a dummy data packet, means, such as data receiver 44 for monitoring channel 16, and data buses 46, 48, and 49.

The Ethernet networks illustrated by FIG. IA and FIG. IB are functionally equivalent to each other in that nodes 10, 12, 14, and 15 communicate with each other using the same Ethernet protocol. The implementational difference between the two networks is that in FIG. IA, channel 16 is comprised of a coaxial cable whereas in FIG. IB, channels 16 are comprised of pairs of twisted-pair wires, both implementations being well-known in the art. Aβ is further well-known in the art, a twisted-pair Ethernet network requires that all communications between nodes be routed through hub node 17, aβ illustrated in FIG. IB. The present invention is equally applicable to both Ethernet network implementations, as well as other networks employing other connection media.

Referring to Figures 2 aNd 3, during a node's transmitting and receiving functions, buffer 18 operates to temporarily store data packets pending commencement of transmission across channel 16, and receive buffer 20 operates to temporarily store data packets received from channel 16. MAC 26, in response to CPU 24, may activate transmit status signal 32 whenever one or more data packets are stored in buffer 18 and channel 16 is not busy. When transmit status signal 32 is activated, data packets are transferred from transmit buffer 18 to transmitter 40 via data bus 36 and acroββ channel 16 via data bus 49. Similarly, whenever data packet traffic exists on channel 16 (regardless of the source address and target address of the data packets on the channel), data receiver 44 is operative to receive such incoming data packets via data bus 46, and is particularly provided for activating receive status signal 34 in response thereto. The incoming data packets are transmitted from data receiver 44 to receive buffer 20 via data bus 30. Since standard Ethernet protocol prohibits a node from attempting to transmit data (by activating its transmit statue signal 32) while channel 16 is busy (and the same node's receive statue signal 34 is active), a collision can occur on channel 16 if and only if transmit statue signal 32 and receive statue signal 34 of the same node, for example node 15, are activated substantially simultaneously by way of node 15 and a second node, for example node 12, attempting commencement of transmission at substantially the same time. According to the present invention, jam control signal 38 iβ overlayed onto the standard Ethernet protocol at the physical layer of node 15. JAM control signal 38 is activated directly by CPU 24. When jam control signal 38 is activated, node 15 enters "jam" mode, as shown by steps 62 and 64 illustrated in FIG. 4. While in "jam" mode, node 15 refuses acceptance of incoming data packets by forcing collisione on channel 16. For example, if jam control βignal 38 of node 12, the target node, iβ activated, node 12 enters "jam" mode. According to standard Ethernet protocol, if node 14, the source node, then attempts commencement of tranβmission of data packets across channel 16 to node 12, receive status signal 34 of node 12 (and every other node on the network) becomee activated, as shown in steps 66 and 68 of FIG. 4. When jam control signal 38 and receive status signal 34 are activated contemporaneously, jam transmitter 42 causes a dummy data packet to be transmitted across channel 16 by node 12, as shown in step 70 of FIG. 4. The dummy data packet consists of a legitimate start-of-packet delimiter, followed by a predetermined, fixed data pattern which is generated by jam transmitter 42 at the physical layer of the Ethernet protocol. The content of the fixed data pattern is not critical for purposes of this invention. Source node 14 senses that target node 12 iβ transmitting data on channel 16, and receive status signal 34 of source node 14 is activated. Since transmit statue βignal 32 of source node 14 is activated (due to node 1 's commencement of data packet transmission) and receive status signal 34 of source node 14 iβ also activated, source node 14 senses a collision on channel 16 and ceases transmission of any further data packets. According to standard Ethernet protocol, when source node 14 senses the collision forced by target node 12, source node 14 "backs off" from transmitting and retries itβ transmission of the same data packet at a later random time as calculated by the IEEE 802.3 standard algorithm known as "truncated binary exponential backoff." Thuβ, target node 12 iβ able to effectively control the flow of incoming data packets at the physical layer of the protocol without losing any data packets and without having to resort to the "recovery" procesβ at an upper layer above the Ethernet protocol.

For aβ long as node 12 remains in "jam" mode, node 12 will force collisione as described above with any incoming data packets from channel 16, aβ shown in stepβ 64 through 74 of

FIG. 4. Even though jam control signal 38 of node 12 is activated, node 12 may still transmit data packets acroββ channel 16 aa normal, except while itβ receive status aignal 34 ie alao activated. When node 12 determines that internal node conditiona make it possible or optimal to receive data from channel 16, node 12 de-activates jam control signal 38. Once jam control βignal 38 is de-activated, node 12 no longer forces collisione on channel 16, and node 12 acceptβ incoming data packets from channel 16.

It should be appreciated that jam control signal 38 and the associated method of forcing collisions may be implemented in Ethernet networks incorporating a plurality of nodes. When a node is in "jam" mode and it transmits a dummy data packet, the duration of such transmission lasts a fixed amount of time, which iβ long enough to guarantee collision detection by all nodes throughout the network. In the constructed embodiment, it has been found that a time period correeponding to the tranβmiββion of a 64 data byte frame (minimum IEEE 802.3 frame size) iβ sufficient for all nodeβ to detect the forced collision. Jam control signal 38 is activated by a source external to the physical layer of the Ethernet protocol, such as a central processing unit (CPU) 24, which detects whether internal node conditions make it necessary or optimal for the target node to delay acceptance of incoming data packets. The assertion of jam control signal 38 may depend upon the existence of any number of certain node conditions, such as a utilization level of receive buffer 20. Reference is not made to the characteristics of the channel 16 when deciding to enter "jam" mode, or to the character or nature of the data packets on the channel. In the constructed embodiment, memory 22, in communication with CPU 24, stores βoftware adapted to control CPU 24. The βoftware, in connection with CPU 24, provide the means for determining when the measured utilization level of receive buffer 20 exceeds a predetermined threshold utilization level. The βoftware performs the functions illustrated in FIG. 5. First, the utilization level of receive buffer 20 ie meaβured, etep 76. Second, the measured utilization level of receive buffer 20 iβ compared to a predetermined threshold utilization level, step 78. Finally, if the measured level exceeds the threshold level, the software causes jam control signal 38 to be activated, step 80.

By way of example only, the present invention may be implemented in a network hub, as shown in FIG. IB. When receive buffer 20 of network hub 17 becomes full and unable to accept further incoming data, network hub 17 may activate jam control signal 38, thus causing hub 17 to enter "jam" mode. While in "jam" mode, incoming data will be rejected by hub 17, since hub 17 will be forcing collisione on channel 16. Instead of being "lost", aβ according to standard Ethernet protocol, βuch rejected data packets will be retransmitted by the eource node at a later random time. The preceding description is exemplary rather than limiting in nature. A preferred embodiment of thi^β invention haβ been disclosed to enable one skilled in the art to practice this invention. Variations and modifications are poββible without departing from the spirit and purview of this invention, the ecope of which ie limited only by the appended claims.

Claims

CLAIMSWhat is claimed is:

1. For use in a target node of a network employing a standard CSMA/CD Ethernet protocol for channel acceβe, a method of controlling receipt of incoming data packets by the target node, comprising the steps of: determining, without reference to conditions associated with said channel, when jam criteria are satisfied and activating a jam signal in response thereto; monitoring said channel to determine when said target node is receiving incoming data packets from a source node and activating a receive statue signal in response thereto; transmitting a dummy data packet from said target node to said channel when said jam signal and said receive status signal are activated contemporaneously to thereby force a collision wherein said source node, in accordance with said standard protocol, will defer further attempts to access said channel.

2. The method of claim 1, further comprising the step of: repeating said monitoring etep and βaid transmitting step until said jam criteria cease to be satisfied and deactivating said jam signal in response thereto.

3. The method of claim 2, wherein said step of transmitting a dummy data packet from said target node to said channel is performed automatically at the physical layer of the CSMA/CD protocol, bypassing standard Ethernet transmission method of submitting said dummy data packet to a Media Access Controller (MAC) for transmission.

4. The method of claim 3, wherein said determining step is performed as a function of a buffer utilization level.

5. The method of claim 4, wherein said determining step comprises the substeps of: measuring a utilization level of a data receive buffer of said target node; comparing said measured utilization level with a predetermined threβhold utilization level; and activating βaid jam βignal when βaid meaβured utilization level exceeda said threshold utilization level.

6. For use in a target node of a network employing a standard CSMA/CD Ethernet protocol for channel access, an apparatus for controlling receipt of incoming data packets by said target node, comprising: means for monitoring said channel to determine when βaid target node iβ receiving incoming data packets from a source node and for activating a receive status signal in response thereto; and means responsive to a jam signal that is indicative of a condition when jam criteria are satisfied and further reβponsive to said receive βtatus aignal for transmitting a dummy data packet from said target node to said channel when said jam signal and said receive statue βignal are activated contemporaneously, thereby forcing a collision on said channel wherein βaid aource node, in accordance with βaid standard protocol, will defer further attempts to access said channel.

7. An apparatus aβ in claim 6, wherein aaid transmitting means compriβeβ a jam transmitter coupled to said channel for transmitting said dummy data packet, wherein said dummy data packet is of a predetermined, fixed format.

8. An apparatus as in claim 6, further comprising: means for determining, without reference to conditions aseociated with said channel, when said jam criteria are satiefied and for activating aaid jam signal in response thereto.

9. An apparatus as in claim 8, wherein said determining means comprises: a controller; a memory in communication with said controller for storing data wherein said data includes at least one βoftware program for controlling said controller.

10. An apparatus aβ in claim 9, wherein βaid monitoring means comprises a data receiver in communication with said channel.

11. An apparatus as in claim 10, further comprising a data receive buffer in communication with said CPU and βaid data receiver.

12. An apparatus as in claim 8, wherein said determining means includes means for determining when an actual receiving buffer utilization level exceeds a predetermined maximum buffer utilization level.

13. For use in a target node of a network employing a etandard CSMA/CD Ethernet protocol for channel access, an apparatus for controlling receipt of incoming data packets by said target node, comprising: 5 a data receiver in communication with said channel for monitoring said channel to determine when said target node iβ receiving incoming data packets from a aource node and for activating a receive βtatuβ βignal in response thereto; a jam transmitter responsive to a jam signal --0 indicative of a condition when jam criteria are βatisfied and said receive status signal for transmitting a dummy data packet from said target node to said channel when said jam signal and said receive status signal are activated contemporaneously, thereby forcing a collision on said channel wherein said source 15 node, in accordance with βaid etandard protocol, will defer further attempts to access said channel. a data transmitter responsive to a transmit status signal and in communication with said channel for transmitting data to said channel; 20 a receive data buffer in communication with said controller and said data receiver for temporarily storing incoming data packets received from said channel; a transmit data buffer in communication with said controller and said data transmitter for temporarily storing 25 outgoing data packeta pending transmission to said channel; a controller in communication with said transmit data buffer and said receive data buffer and providing said transmit status signal to said data transmitter and further providing said jam signal to said jam transmitter; and 30 a memory in communication with aaid controller for storing data wherein said data includes at least one software program for controlling said controller.