Advanced Micro Devices Corporation

Jayant Kadambi

Mohan V Kalkunte

Jim Mangin

Shashank C Merchant

Thomas J Runaldue

Bahadir Erimli

Ian S Crayford

Denise Kerstein

Philip Simmons

Jinqlih Sang

Peter Ka-Fai Chow

Gopal S Krishna

Eric Tsin-Ho Leung

James Lawrence Mangin

Ganatios Y Hanna

Apparatus and method for more precise controlling of congestion on a network, provides for remote controlling of a remote station on the network by a local station to configure the remote station into a remote loopback configuration. With the remote station thus configured, the local station is then able to determine the link latency of the link, during auto-negotiation, for example. Provided with the link latency, a congestion control algorithm in the local station may be adjusted to account for the link latency to better control the input data streams by controlling when the congestion relieving control signal, such as a PAUSE frame, is transmitted to the remote station to inhibit transmission and relieve congestion.

Arrangement for determining link latency for maintaining flow control in full-duplex networks

A network switch in a packet switched network has multiple network ports, each having a physical layer transceiver, a media access controller, transmit and receive buffers, and a bus interface for transferring data using a host processor. The network switch includes a low-latency mode, where a transfer request is initiated by a network switch port when the receive buffer includes 16 bytes of packet data. If during transfer the MAC detects a collision on the shared network medium, the network port generates a HALT signal in response to detecting a collision, enabling the host controller transferring data from the network port to abort the transfer, and begin transfer on another port. Status information also is stored in the receive buffer in response to the detected collision. Generation of the HALT signal enables the network switch to operate in low-latency mode without requiring substantial resources by the host processor to filter collision fragments or runt packets.

Apparatus and method for terminating a data transfer in a network switch in response to a detected collision

A network switch having a shared memory architecture for storing data frames has a set of programmable thresholds that specify when flow control should be initiated on a selected network port. The network switch includes a queue for storing free frame pointers, each specifying available memory locations in an external memory for storing data frames received from a network station. The network switch takes a frame pointer from a free buffer queue for each received data frame, and stores the received data frame in the location in external memory specified by the frame pointer while a decision making engine within the switch determines the appropriate destination ports. Flow control is initiated on a half-duplex network port by transmitting a phantom packet to a transmitting network station if the output buffer of a destination port exceeds a programmed high threshold, or if the output buffer of the destination port exceeds a low programmed threshold and the queue of free frame pointers falls below a low programmable threshold. The switch thus provides flexibility in providing selective collisions with transmitting stations to minimize the possibility of network congestion without wasting network bandwidth with unnecessary collisions.

Method and apparatus providing programmable thresholds for half-duplex flow control in a network switch

A network having a shared memory architecture for storing data frames has a set of programmable thresholds that specify when flow control should be initiated on full-duplex network ports. The network switch includes a queue for storing free frame pointers that specify available memory locations in an external memory for storing data frames received from a network station. The network switch takes a frame pointer from a free buffer queue for each received data frame, and stores the received data frame in the location in external memory specified by the frame pointer while a decision making engine within the switch determines the appropriate destination ports. Flow control is initiated based on the number of available frame pointers by transmitting a PAUSE frame having a selected PAUSE interval to a transmitting network station. Specifically, a full-duplex port will output a PAUSE frame having a short, medium, or long programmed pause interval if the free buffer pool of available frame pointers falls below a high, medium, or low programmable threshold, respectively. The switch thus provides flexibility in generating variable-length PAUSE control frames to minimize wasting network bandwidth.

Method and apparatus providing programmable thresholds for full-duplex

A network switch in a full-duplex IEEE 802.3 network includes a data monitor module that monitors data utilization between ports. Upon detection of a congestion condition in an output buffer, the data monitor module determines for each of the remaining network switch ports a traffic contribution relative to the total network traffic received by the one congested network switch port. The network switch includes a rate controller that generates rate control frames for the remaining network switch ports, where each rate control frame specifies a transmission rate for the transmitting network node based on the corresponding traffic contribution by the network node. The outputting of a rate control frame to a network node based on the corresponding traffic contribution optimizes elimination of the congestion condition by prioritizing the generation of rate control frames for the network node most responsible for creating the congestion condition. In addition, disruption of traffic is minimized in stations having little or no traffic contribution to the congested network switch port.

Apparatus and method for generating rate control frames in a workgroup

A network interface for a shared gigabit Ethernet network selectively modulates an interpacket gap interval following a burst transmission in order to establish a rotating priority arrangement with network stations on the gigabit network. A network station includes a programmable burst timer that counts a burst interval corresponding to a negotiated bandwidth. The network station having accessed the media continues to transmit data packets so long as data is available in a transmit buffer, and the burst timer has not expired. Each data packet within the burst is transmitted after waiting a minimum interpacket gap interval of 96 bit times. Following the burst transmission, the network interface waits a modified delay interval equal to the minimum interpacket gap plus a multiple number of slot times related to the number of stations on the network. The modified delay interval is decremented by a slot time each time the network station detects a burst transmission by another network station. Each network station thus transmits a burst of data packets according to a negotiated bandwidth, and minimizes the number of encountered collisions by deferring to other network stations following a burst transmission.

Apparatus and method for regulating assigned bandwidth in high speed

A network switch in a full-duplex IEEE 802.3 network includes a data monitor module that monitors data utilization between ports. Upon detection of a congestion condition in an output buffer, the data monitor module determines for each of the remaining network switch ports a traffic contribution relative to the total network traffic received by the one congested network switch port. The network switch includes a pause controller that generates pause control frames for the remaining network switch ports, where each pause control frame specifies a corresponding interval having a duration based on the traffic contribution by the corresponding network node. The outputting of a pause control frame to a network node based on the corresponding traffic contribution optimizes elimination of the congestion condition by prioritizing the generation of pause frames for the network node most responsible for creating the congestion condition. In addition, disruption of traffic is minimized in stations having little or no traffic contribution to the congested network switch port.

Apparatus and method for generating flow control frames in a workgroup

A network switch having switch ports for full-duplex communication of data packets with respective network nodes according to IEEE 802.3 protocol dynamically allocates bandwidth between the switch ports based upon detected activity from the network nodes. The network switch generates an assigned bandwidth value for each active switch port based upon the switch capacity and the number of active switch ports. Each active switch port forwards the assigned bandwidth value to the corresponding network node as an IEEE 802.3x[2] compliant media access control (MAC) control frame, enabling the corresponding network node to calculate its programmed interpacket gap interval following a packet transmission based upon the size of the transmitted packet and the assigned bandwidth. Each active switch port also calculates the programmed interpacket gap to determine if reception of another data packet has begun by the end of the programmed interpacket gap interval. If no data packet is received by the corresponding port after the programmed interpacket gap interval, the network switch considers that switch port as non-active, recalculates, and reassigns the unused bandwidth to the remaining active ports. The dynamic bandwidth allocation by the network switch enables the input buffer size to be optimized without requiring an excessive switch bandwidth.

Apparatus and method in a network switch for dynamically allocating

The present invention comprises a system and method for reducing the pin count between a plurality of MAC and PHY devices within a switching element. In this embodiment, the switching element includes a plurality of general serial interfaces for providing connections between respective MAC and PHY devices and each of the plurality of general serial interfaces operates at a first data rate. The system and method comprises a multiplexer coupled to the plurality of general serial interfaces and a pad member including a plurality of pins. The pad member is coupled to the multiplexer and receives multiplexed signals from the plurality of general serial interfaces. The multiplexer operates at a second data rate that is a multiple of the first data rate. Generally, a system and method in accordance with the present invention allows for the multiplexing of a general purpose serial interface (GPSI) to reduce the pin count in some cases by as much as 75% and also synchronize the MAC/PHY interface. In this example, the multiplexer interface uses a total of 7 pins and supports a total of four MAC/PHY connections. If only GPSIs were utilized, 28 pins would be required for this function. The same multiplexing technique will also reduce the MAC/PHY interface in four 100 Mbps connections from 56 pins for a four port system to 18 pins. In each example the multiplexer interface will operate at four times the speed of the general serial interface.

Reducing the pin count within a switching element through the use of a

A buffered distributor (i.e., a full-duplex repeater) having receive buffers for respective network ports calculates pause frames based on the size of stored data packets that need to be output by the repeater core according to a round robin sequence before congestion in an identified receive buffer is eliminated. The distribution core within the buffered distributor includes a linked list register that stores the determined links of received data packets for each network port. Upon detecting a congestion condition in one of the receive buffers for a corresponding port, the buffered distributor determines the relative position of the congested port within the round robin sequence, and calculates the pause interval based on the length of the data packets that need to be output before congestion is eliminated. The sum of the data packet lengths are compared to an output data rate of the distribution core, as well as switching delays within the core. The calculated pause frames, in conjunction with the prescribed congestion threshold, ensures that buffer sizes can be efficiently designed at a low cost, without compromising cost or risk of lost data packets.

Apparatus and method for generating a pause frame in a buffered

Interpacket delay times are modified in full-duplex Ethernet network devices by calculating for each network station a delay interval based on a time to transmit a data packet at the network rate and a calculated time to transmit the data packet at a desired transmission rate. The network station waits the calculated delay time following a packet transmission before transmitting the next data packet, ensuring that the overall output transmission rate of the network station corresponds to the assigned desired transmission rate. The desired transmission rate is received as a media access control (MAC) control frame from a network management entity, such as a switched hub. Hence, each network station operates at the desired transmission rate, minimizing the occurrence of congestion and eliminating the necessity of PAUSE frames.

Method and apparatus for regulating data flow in networks

Delay times are modified in full-duplex Ethernet network devices by calculating in each network station a delay interval based on a size of a transmitted data packet and a desired transmission rate. The network station waits the calculated delay time before transmitting another data packet, ensuring that the overall output transmission rate of the network station corresponds to the assigned desired transmission rate. The desired transmission rate is received as a media access control (MAC) control frame from a network management entity, such as a switched hub. Hence, each station operates at the desired transmission rate, minimizing the occurrence of congestion and eliminating the necessity of PAUSE frames.

Arrangement for regulating packet flow rate in shared-medium,

Collision delay intervals are modified in Ethernet network devices transmitting priority data requiring a guaranteed latency by multiplying an integer multiple number of slot times with a fractional coefficient. A network device having priority data for transmission uses the conventional Truncated Binary Exponential Backoff (TBEB) algorithm during the first access attempt to calculate a collision delay interval from a randomly selected integer multiple of slot times. If the network device encounters another collision, the next randomly selected integer multiple of slot times is multiplied by the fractional coefficient. Use of the fractional coefficient during collision mediation on a half-duplex Ethernet network provides a bounded access latency for real-time and multimedia applications by granting the network device a higher probability of successfully accessing the network media.

Method and apparatus for prioritizing traffic in half-duplex networks

Delay times are modified in Ethernet network devices by adding an integer multiple of a delay interval to the minimum interpacket gap (IPG) interval, and decrementing a deferral counter storing the integer in each network station in response to detected activity on the media. Each station independently determines the number of stations active on the network media by counting the number of successful packet receptions following a corresponding detected collision. Once the number of detected collisions equals the number of stations (N) minus one, each station independently establishes a unique integer value from the range of zero to the number of detected collisions, i.e., up to the number of stations (N) minus one, by resetting the deferral counter to (N-1) after a successful transmission, and by decrementing the deferral counter upon detection of a successful transmission without collision by another station. The unique integer value ensures that each station has a different delay interval in accessing the media after sensing deassertion of the receive carrier. Each network station also includes a deferral timer that counts the maximum delay interval of (N-1) delay intervals plus the minimum IPG value, and thus establishes a bounded access latency for a half-duplex shared network.

Determining the number of active nudes on an ethernet network by

Efficient communication in a network having a minimum data transmission time interval wherein a data packet is transmitted beginning at a start of the minimum data transmission time interval. An end of the minimum data transmission time interval is determined and then at least one next data packet is transmitted after the data packet if the end of the minimum data transmission time is after the end of the data packet.

Growing community of inventors

Milpitas, CA, United States of America

Jayant Kadambi