Advanced Micro Devices Corporation

Yatin Acharya

Bahadir Erimli

Shr-Jie Tzeng

Peter Ka-Fai Chow

Matthew James Fischer

Philip J Keller

Oikwan Tsang

Somnath Viswanath

Atul Garg

Hung Duy Vo

A network manager, configuring for detecting network nodes and configuring network switches, determines addressing field lengths to be used for addressing the network nodes and switching data packets between the network nodes based on the number of detected network nodes. The network manager detects the network nodes by exploring the network according to prescribed explorer procedures. The network manager selects a size of address fields to be used for switching data packets traversing the network, based on the number of detected network nodes. The network manager configures each network switch within the network to switch the data packets based on a switching tag having the selected size and positioned at the start of the packet. Hence, each network switch is able to generate forwarding decisions based on the switching tag at the beginning of each received data packet. The switching tag is distinct from, and substantially smaller than, the existing destination address field. Hence, switching complexity can be minimized for relatively small networks having minimal addressing requirements, reducing latency and simplifying forwarding decisions within the network switches.

Arrangement for switching infiniband packets using switching tag at start of packet

An InfiniBand&#x2122; network node includes a network interface, a system memory, a memory controller configured for controlling access to the system memory, and a processor. The network interface is configured for outputting data packets according to a prescribed flow control protocol that specifies flow control resources. The network interface also is configured for outputting a data flow interruption request to the memory controller based on a determined depletion of the flow control resources. The memory controller, in response to reception of the data flow interruption request, restricts access to the system memory. Hence, the processor, in response to detecting the restricted access to the system memory, reduces execution of a prescribed application resource based on the determined depletion of the flow control resources. Hence, the supply of data for the data packets is reduced based on the reduced execution of the prescribed application resource, enabling flow control protocols to be maintained while preserving processor and memory resources for other application resources.

Arrangement for reducing application execution based on a determined lack of flow control credits for a network channel

A network device identifies priority level information for data frames it receives. The network device includes input ports, a memory, an action generator, and a port vector queue. The input ports receive the data frames. Each of the received data frames specifies one or more classes of service. The memory stores priority level information corresponding to each of the classes of service. The action generator generates an action tag for each of the received data frames. The port vector queue uses the action tag from the action generator for each of the received data frames to access the memory to identify the priority level information associated with the received data frame.

Systems and methods for expediting the identification of priority information for received packets

A network device includes a group of queues, each having a weighted round robin mechanism. The priority queues on a port detect an overflow condition and transfer a flag to the weighted round robin device in response to detecting the overflow condition. The weighted round robin mechanism adjusts the weight associated with one or more of the priority queues in response to receiving the flag and transfers data from the queues based on the adjusted weights.

System and method for dynamically updating weights of weighted round robin in output queues

A host channel adapter configured for outputting packets, according to a service protocol requiring acknowledgement messages within a prescribed time interval following transmission, utilizes a retransmission table for storing entries identifying the packets awaiting respective acknowledgment messages during the respective prescribed time intervals. A retransmission manager is configured for updating the retransmission table after each access cycle, defined as a prescribed number of clock cycles. The retransmission manager also identifies a number of transmitted packets within the corresponding access cycle within a selected initial entry for the access cycle. An acknowledgment manager in the receive portion of the host channel adapter resets to zero an acknowledgment waiting bit in a selected entry in response to an acknowledgment message identifying the corresponding packet. If the retransmission manager determines after a prescribed retransmission time that the selected initial entry for the access cycle does not reset an acknowledgment waiting bit, the entries identifying the transmitted packets within the corresponding access cycle are transferred to a transmit queue for retransmission.

Arrangement for managing transmitted packets requiring acknowledgement in a host channel adapter

A network device determines forwarding information for received data frames. The network device includes input ports, queuing logic, a forwarding engine, and a port filter. The input ports receive data frames. The queuing logic transfers at least some of the received data frames to an external memory. The forwarding engine generates forwarding information for at least some of the data frames transferred by the queuing logic to the external memory. The port filter stores forwarding information for one or more of the received data frames and analyzes each of the received data frames to determine whether there is stored forwarding information related to the received data frame. When there is stored forwarding information related to the received data frame, the port filter uses the stored forwarding information to forward the received data frame.

Systems and methods for bypassing packet lookups

An ATM&#x2014;InfiniBand&#x2122; router is configured for interfacing between an asynchronous transmission mode (ATM) network and an InfiniBand&#x2122; network, without a necessity of intermediate transport on a packet based network such as an Internet Protocol (IP) network. The router includes an ATM processor and a host channel adapter. The ATM processor is configured for generating ATM cells streams based on received InfiniBand&#x2122; packets, and recovering InfiniBand&#x2122; packet data from received ATM cells. The host channel adapter is configured for receiving the InfiniBand&#x2122; packets from the InfiniBand network and providing at least the payload data to the ATM processor, and outputting the recovered InfiniBand&#x2122; packet data onto the InfiniBand&#x2122; network. In addition, the ATM processor and the host channel adapter may be configured for mapping the ATM cells and the InfiniBand packets on prescribed virtual circuits and prescribed InfiniBand&#x2122; connections, respectively. Hence, the ATM&#x2014;InfiniBand&#x2122; router operates as a call connection handler, enabling connections to be established across ATM and InfiniBand&#x2122; networks.

Arrangement for converting ATM cells to infiniband packets

A system automatically establishes a trunk between first and second network devices. The system monitors a source address and destination addresses in packets destined to or received from the second network device and determines whether a communication pattern exists. When a communication pattern is determined to exist, the system automatically establishes the trunk between the first network device and second network device.

Self-configuring trunking on a network device

A router is configured for sending and receiving data packets on an InfiniBand&#x2122; network. When placed between an Ethernet network and an InfiniBand&#x2122; network, the router is configured to receive an Ethernet data packet having a VLAN tag indicative of layer 2 priority data of the Ethernet packet. The router includes a mapping table having multiple entries, each entry specifying a VLAN tag and a corresponding service level. A controller is configured for parsing the VLAN tag and determining the service level for the VLAN tag. The controller outputs the Ethernet packet on the InfiniBand&#x2122; network within an InfiniBand&#x2122; packet according to the determined service level.

Mapping layer 2 LAN priorities to a virtual lane in an Infiniband&#x2122; network

A network device includes a port filter, a first logic device, and a second logic device. The port filter receives a data frame and generates first data relating to the data frame. The first logic device generates second data for the received data frame. The second logic device receives the first data and the second data, determines whether the first data contains a valid first priority value, and assigns the valid first priority value to the data frame when the first data contains the valid first priority value. When the first data does not contain a valid first priority value, the second logic device determines whether the second data contains a valid second priority value, and assigns the valid second priority value to the data frame when the second data contains the valid second priority value.

System and method for deciding outgoing priority for data frames

A network device includes a filter and an embedded processor. The filter receives a packet, determines whether the packet is an RSVP packet, and transmits an interrupt signal when the packet is an RSVP packet. The embedded processor receives the interrupt signal and generates at least one packet processing parameter in response to the interrupt signal.

Intelligent embedded processor enabled mechanism to implement RSVP function

A network interface includes a network medium interface operatively coupled to a software device driver arrangement, with a legacy media access controller (MAC) therebetween. The device driver arrangement includes a legacy MAC device driver configured to communicate with the legacy MAC, and an intermediate driver configured to communicate with the network medium interface. The intermediate driver and the network medium interface may communicate with one another by use of special frames, for example to send and receive control information. The special frames are formatted to pass through the legacy MAC, and include an identifier so that they can be identified at the intended destination, either the intermediate driver or the network medium interface. Upon identification, the control information is extracted at the destination. The network medium interface may include one or more MACs as well as one or more physical layer devices (PHYs).

Network interface apparatus and method of internal communication within the same

A router is configured for sending and receiving data packets on an InfiniBand&#x2122; network. The router is configured to receive a network layer data packet having a transport header having an application identifier indicative of application layer priority data of the network layer packet. The router includes a mapping table having multiple entries, each entry specifying an application identifier and a corresponding service level. A controller is configured for parsing the transport header and determining the service level for the application identifier. The controller outputs the network layer packet on the InfiniBand&#x2122; network within an InfiniBand&#x2122; packet according to the determined service level.

Using application headers to determine InfiniBand&#x2122; priorities in an InfiniBand&#x2122; network

A multiport network device includes output port logic, priority logic, a memory, and memory logic. The output port logic generates output port data that identifies output ports to transmit received packets. The priority logic generates priority data that identifies priorities of the received packets. The memory stores the output port data from the output port logic and the priority data from the priority logic. The memory logic receives priority data relating to one of the received packets from the output port logic, determines whether the memory stores output port data relating to the packet, ignores the received priority data when the memory stores no output port data relating to the packet, and when the memory stores output port data relating to the packet, transmits the received priority data and the stored output port data to the identified output port.

Systems and methods for merging packet lookup results

A network device that controls the communication of data frames between stations includes a number of receive ports that receive data frames from the stations and a number of output ports that transmit the data frames to their intended destinations. The network device also includes a number of output queues that store data forwarding information associated with the received data frames. The network device partitions each of the output queues into a number of portions corresponding to the priorities supported by the network device. The number of portions and the size of each portion of the output queues may be programmable by the user.

Growing community of inventors

Sunnyvale, CA, United States of America

Yatin Acharya