Receiver-based precision congestion control
Examples described herein relate to a network agent, when operational, to: receive a packet, determine transmit rate-related information for a sender network device based at least on operational and telemetry information accumulated in the received packet, and transmit the transmit rate-related information to the sender network device. In some examples, the network agent includes a network device coupled to a server, a server, or a network device. In some examples, the operational and telemetry information comprises: telemetry information generated by at least one network device in a path from the sender network device to the network agent.
1. Integrated circuit for use in association with ternary content-addressable memory (TCAM) in a network switch, the network switch being configurable for use in a network that comprises multiple server nodes and multiple switch nodes, the multiple server nodes comprising at least one server node and at least one other server node, multiple switch nodes comprising at least one switch node and at least one other switch node, the network switch to be comprised in the at least one switch node, the integrated circuit comprising:
packet processing pipeline circuitry configurable to implement ingress and egress pipelines, the ingress and egress pipelines comprising at least one parser stage, multiple match-action stages, and at least one other stage, the at least one parser stage to parse packet header data, the multiple match-action stages to match parsed header field data with at least one portion of table data stored, at least in part, in the TCAM to determine packet processing action data associated with the parsed header field data, the at least one other stage to generate output packet data for output from the integrated circuit; and
traffic manager circuitry to queue ingress pipeline processed packet data for use by at least one egress pipeline;
wherein:
the integrated circuit is to generate telemetry-related information for use in determining packet flow transmit-rate information for at least one packet flow transmitted from the at least one server node in the network;
the packet flow transmit-rate information is to be used by the at least one server node to adjust flow-rate of the at least one packet flow from the at least one server node in the network;
the telemetry-related information is to be accumulated, with other telemetry-related information from the at least one other switch node, to generate accumulated telemetry-related information;
the accumulated telemetry-related information is (1) to be provided to the at least one other server node and (2) for use in generating the packet flow transmit-rate information by the at least one other server node;
the accumulated telemetry-related information comprises:
timestamp information of the at least one switch node and the at least one other switch node;
queue depth information of the at least one switch node and the at least one other switch node; and
bytes transmitted information of the at least one switch node and the at least one other switch node; and
the flow-rate of the at least one packet flow from the at least one server node in the network is to be adjusted, based upon the packet flow transmit-rate information, in association with implementing network congestion control associated with the at least one packet flow.
2. The integrated circuit of claim 1 , wherein:
the at least one packet flow is to be identified based upon protocol tuple information; and
the protocol tuple information comprises packet header information.
3. The integrated circuit of claim 2 , wherein:
at least one other portion of the table data comprises exact match table data.
4. The integrated circuit of claim 3 , wherein:
the accumulated telemetry-related information is to be provided, via acknowledge (ACK) packet data, to the at least one server node from the at least one other server node.
5. The integrated circuit of claim 4 , wherein:
the flow-rate of the at least one packet flow from the at least one server node in the network comprises a remote direct memory access (RDMA) over converged Ethernet (RoCE) traffic flow rate;
the network congestion control is compatible with High Precision Congestion Control (HPCC); and
the telemetry-related information that is to be generated by the integrated circuit, the other telemetry-related information, and/or the accumulated telemetry-related information are compatible with Internet Engineering Task Force (IETF) Inband Flow Analyzer (IFA) telemetry information.
6. The integrated circuit of claim 4 , wherein:
the integrated circuit is comprised in an application specific integrated circuit;
the packet processing pipeline circuitry comprises a data plane; and
the packet processing pipeline circuitry and the data plane are programmable using Network Programming Language (NPL).
7. The integrated circuit of claim 4 , wherein:
the integrated circuit is to generate the accumulated telemetry-related information; and
the at least one switch node is to provide the accumulated telemetry-related information to the at least one other server node.
8. A method implemented using an integrated circuit, the integrated circuit to be used in association with ternary content-addressable memory (TCAM) in a network switch, the network switch being configurable for use in a network that comprises multiple server nodes and multiple switch nodes, the multiple server nodes comprising at least one server node and at least one other server node, the multiple switch nodes comprising at least one switch node and at least one other switch node, the network switch to be comprised in the at least one switch node, the integrated circuit comprising packet processing pipeline circuitry configurable to implement ingress and egress pipelines, and traffic manager circuitry, the ingress and egress pipelines comprising at least one parser stage, multiple match-action stages, and at least one other stage, the method comprising:
parsing, by the at least one parser stage, packet header data;
matching, by the multiple match-action stages, parsed header field data with at least one portion of table data stored, at least in part, in the TCAM to determine packet processing action data associated with the parsed header field data;
queuing, by the traffic manager circuitry, ingress pipeline processed packet data for use by at least one egress pipeline; and
generating, by the at least one other stage, output packet data for output from the integrated circuit;
wherein:
the integrated circuit is to generate telemetry-related information for use in determining packet flow transmit-rate information for at least one packet flow transmitted from the at least one server node in the network;
the packet flow transmit-rate information is to be used by the at least one server node to adjust flow-rate of the at least one packet flow from the at least one server node in the network;
the telemetry-related information is to be accumulated, with other telemetry-related information from the at least one other switch node, to generate accumulated telemetry-related information;
the accumulated telemetry-related information is (1) to be provided to the at least one other server node and (2) for use in generating the packet flow transmit-rate information by the at least one other server node;
the accumulated telemetry-related information comprises:
timestamp information of the at least one switch node and the at least one other switch node;
queue depth information of the at least one switch node and the at least one other switch node; and
bytes transmitted information of the at least one switch node and the at least one other switch node; and
the flow-rate of the at least one packet flow from the at least one server node in the network is to be adjusted, based upon the packet flow transmit-rate information, in association with implementing network congestion control associated with the at least one packet flow.
9. The method of claim 8 , wherein:
the at least one packet flow is to be identified based upon protocol tuple information; and
the protocol tuple information comprises packet header information.
10. The method of claim 9 , wherein:
at least one other portion of the table data comprises exact match table data.
11. The method of claim 10 , wherein:
the accumulated telemetry-related information is to be provided, via acknowledge (ACK) packet data, to the at least one server node from the at least one other server node.
12. The method of claim 11 , wherein:
the flow-rate of the at least one packet flow from the at least one server node in the network comprises a remote direct memory access (RDMA) over converged Ethernet (RoCE) traffic flow rate;
the network congestion control is compatible with High Precision Congestion Control (HPCC); and
the telemetry-related information that is to be generated by the integrated circuit, the other telemetry-related information, and/or the accumulated telemetry-related information are compatible with Internet Engineering Task Force (IETF) Inband Flow Analyzer (IFA) telemetry information.
13. The method of claim 11 , wherein:
the integrated circuit is comprised in an application specific integrated circuit;
the packet processing pipeline circuitry comprises a data plane; and
the packet processing pipeline circuitry and the data plane are programmable using Network Programming Language (NPL).
14. The method of claim 11 , wherein:
the integrated circuit is to generate the accumulated telemetry-related information; and
the at least one switch node is to provide the accumulated telemetry-related information to the at least one other server node.
15. At least one non-transitory machine-readable storage medium storing instructions for being executed by at least one machine that is associated with an integrated circuit, the integrated circuit to be used in association with ternary content-addressable memory (TCAM) in a network switch, the network switch being configurable for use in a network that comprises multiple server nodes and multiple switch nodes, the multiple server nodes comprising at least one server node and at least one other server node, the multiple switch nodes comprising at least one switch node and at least one other switch node, the network switch to be comprised in the at least one switch node, the integrated circuit comprising packet processing pipeline circuitry configurable to implement ingress and egress pipelines, and traffic manager circuitry, the ingress and egress pipelines comprising at least one parser stage, multiple match-action stages, and at least one other stage, the instructions, when executed by the at least one machine, resulting in performance of operations comprising:
parsing, by the at least one parser stage, packet header data;
matching, by the multiple match-action stages, parsed header field data with at least one portion of table data stored, at least in part, in the TCAM to determine packet processing action data associated with the parsed header field data;
queuing, by the traffic manager circuitry, ingress pipeline processed packet data for use by at least one egress pipeline; and
generating, by the at least one other stage, output packet data for output from the integrated circuit;
wherein:
the integrated circuit is to generate telemetry-related information for use in determining packet flow transmit-rate information for at least one packet flow transmitted from the at least one server node in the network;
the packet flow transmit-rate information is to be used by the at least one server node to adjust flow-rate of the at least one packet flow from the at least one server node in the network;
the telemetry-related information is to be accumulated, with other telemetry-related information from the at least one other switch node, to generate accumulated telemetry-related information;
the accumulated telemetry-related information is (1) to be provided to the at least one other server node and (2) for use in generating the packet flow transmit-rate information by the at least one other server node;
the accumulated telemetry-related information comprises:
timestamp information of the at least one switch node and the at least one other switch node;
queue depth information of the at least one switch node and the at least one other switch node; and
bytes transmitted information of the at least one switch node and the at least one other switch node; and
the flow-rate of the at least one packet flow from the at least one server node in the network is to be adjusted, based upon the packet flow transmit-rate information, in association with implementing network congestion control associated with the at least one packet flow.
16. The at least one non-transitory machine-readable storage medium of claim 15 , wherein:
the at least one packet flow is to be identified based upon protocol tuple information; and
the protocol tuple information comprises packet header information.
17. The at least one non-transitory machine-readable storage medium of claim 16 , wherein:
at least one other portion of the table data comprises exact match table data.
18. The at least one non-transitory machine-readable storage medium of claim 17 , wherein:
the accumulated telemetry-related information is to be provided, via acknowledge (ACK) packet data, to the at least one server node from the at least one other server node.
19. The at least one non-transitory machine-readable storage medium of claim 18 , wherein:
the flow-rate of the at least one packet flow from the at least one server node in the network comprises a remote direct memory access (RDMA) over converged Ethernet (RoCE) traffic flow rate;
the network congestion control is compatible with High Precision Congestion Control (HPCC); and
the telemetry-related information that is to be generated by the integrated circuit, the other telemetry-related information, and/or the accumulated telemetry-related information are compatible with Internet Engineering Task Force (IETF) Inband Flow Analyzer (IFA) telemetry information.
20. The at least one non-transitory machine-readable storage medium of claim 18 , wherein:
the integrated circuit is comprised in an application specific integrated circuit;
the packet processing pipeline circuitry comprises a data plane; and
the packet processing pipeline circuitry and the data plane are programmable using Network Programming Language (NPL).
21. The at least one non-transitory machine-readable storage medium of claim 18 , wherein:
the integrated circuit is to generate the accumulated telemetry-related information; and
the at least one switch node is to provide the accumulated telemetry-related information to the at least one other server node.
22. A network switch configurable for use in a network that comprises multiple server nodes and multiple switch nodes, the multiple server nodes comprising at least one server node and at least one other server node, multiple switch nodes comprising at least one switch node and at least one other switch node, the network switch to be comprised in the at least one switch node, the network switch comprising:
ports; and
an integrated circuit for use in association with ternary content-addressable memory (TCAM), the integrated circuit to be communicatively coupled to the network via the ports, the integrated circuit comprising:
packet processing pipeline circuitry configurable to implement ingress and egress pipelines, the ingress and egress pipelines comprising at least one parser stage, multiple match-action stages, and at least one other stage, the at least one parser stage to parse packet header data, the multiple match-action stages to match parsed header field data with at least one portion of table data stored, at least in part, in the TCAM to determine packet processing action data associated with the parsed header field data, the at least one other stage to generate output packet data for output from the integrated circuit; and
traffic manager circuitry to queue ingress pipeline processed packet data for use by at least one egress pipeline;
wherein:
the integrated circuit is to generate telemetry-related information for use in determining packet flow transmit-rate information for at least one packet flow transmitted from the at least one server node in the network;
the packet flow transmit-rate information is to be used by the at least one server node to adjust flow-rate of the at least one packet flow from the at least one server node in the network;
the telemetry-related information is to be accumulated, with other telemetry-related information from the at least one other switch node, to generate accumulated telemetry-related information;
the accumulated telemetry-related information is (1) to be provided to the at least one other server node and (2) for use in generating the packet flow transmit-rate information by the at least one other server node;
the accumulated telemetry-related information comprises:
timestamp information of the at least one switch node and the at least one other switch node;
queue depth information of the at least one switch node and the at least one other switch node; and
bytes transmitted information of the at least one switch node and the at least one other switch node; and
the flow-rate of the at least one packet flow from the at least one server node in the network is to be adjusted, based upon the packet flow transmit-rate information, in association with implementing network congestion control associated with the at least one packet flow.
23. The network switch of claim 22 , wherein:
the at least one packet flow is to be identified based upon protocol tuple information; and
the protocol tuple information comprises packet header information.
24. The network switch of claim 23 , wherein:
at least one other portion of the table data comprises exact match table data.
25. The network switch of claim 24 , wherein:
the accumulated telemetry-related information is to be provided, via acknowledge (ACK) packet data, to the at least one server node from the at least one other server node.
26. The network switch of claim 25 , wherein:
the flow-rate of the at least one packet flow from the at least one server node in the network comprises a remote direct memory access (RDMA) over converged Ethernet (RoCE) traffic flow rate;
the network congestion control is compatible with High Precision Congestion Control (HPCC); and
the telemetry-related information that is to be generated by the integrated circuit, the other telemetry-related information, and/or the accumulated telemetry-related information are compatible with Internet Engineering Task Force (IETF) Inband Flow Analyzer (IFA) telemetry information.
27. The network switch of claim 25 , wherein:
the integrated circuit is comprised in an application specific integrated circuit;
the packet processing pipeline circuitry comprises a data plane; and
the packet processing pipeline circuitry and the data plane are programmable using Network Programming Language (NPL).
28. The network switch of claim 25 , wherein:
the integrated circuit is to generate the accumulated telemetry-related information; and
the at least one switch node is to provide the accumulated telemetry-related information to the at least one other server node.
29. A data center comprising:
a network comprising multiple server nodes and multiple switch nodes, the multiple server nodes comprising at least one server node and at least one other server node, multiple switch nodes comprising at least one switch node and at least one other switch node, a network switch to be comprised in the at least one switch node, the network switch comprising:
ports; and
an integrated circuit for use in association with ternary content-addressable memory (TCAM), the integrated circuit to be communicatively coupled to the network via the ports, the integrated circuit comprising:
packet processing pipeline circuitry configurable to implement ingress and egress pipelines, the ingress and egress pipelines comprising at least one parser stage, multiple match-action stages, and at least one other stage, the at least one parser stage to parse packet header data, the multiple match-action stages to match parsed header field data with at least one portion of table data stored, at least in part, in the TCAM to determine packet processing action data associated with the parsed header field data, the at least one other stage to generate output packet data for output from the integrated circuit; and
traffic manager circuitry to queue ingress pipeline processed packet data for use by at least one egress pipeline;
wherein:
the at least one server node is to transmit at least one packet flow in the network;
the integrated circuit is to generate telemetry-related information for use in determining packet flow transmit-rate information for the at least one packet flow that is to be transmitted from the at least one server node in the network;
the at least one server node is to adjust, based upon the packet flow transmit-rate information, flow-rate of the at least one packet flow from the at least one server node in the network;
the telemetry-related information is to be accumulated, with other telemetry-related information from the at least one other switch node, to generate accumulated telemetry-related information;
the at least one other server node is (1) to receive the accumulated telemetry-related information and (2) to generate, based upon the accumulated telemetry-related information, the packet flow transmit-rate information;
the accumulated telemetry-related information comprises:
timestamp information of the at least one switch node and the at least one other switch node;
queue depth information of the at least one switch node and the at least one other switch node; and
bytes transmitted information of the at least one switch node and the at least one other switch node; and
the at least one server node is to adjust the flow-rate of the at least one packet flow from the at least one server node, based upon the packet flow transmit-rate information, in association with implementing network congestion control associated with the at least one packet flow.
30. The data center of claim 29 , wherein:
the at least one packet flow is to be identified based upon protocol tuple information; and
the protocol tuple information comprises packet header information.
31. The data center of claim 30 , wherein:
at least one other portion of the table data comprises exact match table data.
32. The data center of claim 31 , wherein:
the accumulated telemetry-related information is to be provided, via acknowledge (ACK) packet data, to the at least one server node from the at least one other server node.
33. The data center of claim 32 , wherein:
the flow-rate of the at least one packet flow from the at least one server node in the network comprises a remote direct memory access (RDMA) over converged Ethernet (RoCE) traffic flow rate;
the network congestion control is compatible with High Precision Congestion Control (HPCC); and
the telemetry-related information that is to be generated by the integrated circuit, the other telemetry-related information, and/or the accumulated telemetry-related information are compatible with Internet Engineering Task Force (IETF) Inband Flow Analyzer (IFA) telemetry information.
34. The data center of claim 32 , wherein:
the integrated circuit is comprised in an application specific integrated circuit;
the packet processing pipeline circuitry comprises a data plane; and
the packet processing pipeline circuitry and the data plane are programmable using Network Programming Language (NPL).
35. The data center of claim 32 , wherein:
the integrated circuit is to generate the accumulated telemetry-related information; and
the at least one switch node is to provide the accumulated telemetry-related information to the at least one other server node.
36. At least one server node to be comprised in a network comprising multiple server nodes and multiple switch nodes, the multiple server nodes comprising the at least one server node and at least one other server node, multiple switch nodes comprising at least one switch node and at least one other switch node, a network switch to be comprised in the at least one switch node, the at least one server node comprising:
ports to communicatively couple the at least one server node to the network; and
processing circuitry to generate, based upon accumulated telemetry-related information, packet flow transmit-rate information;
wherein:
the network switch comprises an integrated circuit for use in association with ternary content-addressable memory (TCAM);
the integrated circuit comprises:
packet processing pipeline circuitry configurable to implement ingress and egress pipelines, the ingress and egress pipelines comprising at least one parser stage, multiple match-action stages, and at least one other stage, the at least one parser stage to parse packet header data, the multiple match-action stages to match parsed header field data with at least one portion of table data stored, at least in part, in the TCAM to determine packet processing action data associated with the parsed header field data, the at least one other stage to generate output packet data for output from the integrated circuit; and
traffic manager circuitry to queue ingress pipeline processed packet data for use by at least one egress pipeline;
the at least one other server node is to transmit at least one packet flow in the network;
the integrated circuit is to generate telemetry-related information for use in determining the packet flow transmit-rate information for the at least one packet flow that is to be transmitted from the at least one other server node in the network;
the at least one other server node is to adjust, based upon the packet flow transmit-rate information, flow-rate of the at least one packet flow from the at least one other server node in the network;
the accumulated telemetry-related information comprises:
timestamp information of the at least one switch node and the at least one other switch node;
queue depth information of the at least one switch node and the at least one other switch node; and
bytes transmitted information of the at least one switch node and the at least one other switch node; and
the at least one other server node is to adjust the flow-rate of the at least one packet flow from the at least one other server node, based upon the packet flow transmit-rate information, in association with implementing network congestion control associated with the at least one packet flow.
37. The at least one server node of claim 36 , wherein:
the packet flow transmit-rate information is to be communicated to the at least one other server node via remote direct memory access (RDMA) over converged Ethernet (RoCE) protocol;
the at least one packet flow is to be identified based upon protocol tuple information; and
the protocol tuple information comprises packet header information.
38. The at least one server node of claim 37 , wherein:
at least one other portion of the table data comprises exact match table data.
39. The at least one server node of claim 38 , wherein:
the accumulated telemetry-related information is to be provided, via acknowledge (ACK) packet data, to the at least one other server node from the at least one server node.
40. The at least one server node of claim 39 , wherein:
the flow-rate of the at least one packet flow from the at least one other server node in the network comprises at least one ROCE traffic flow rate;
the network congestion control is compatible with High Precision Congestion Control (HPCC); and
the telemetry-related information that is to be generated by the integrated circuit, the other telemetry-related information, and/or the accumulated telemetry-related information are compatible with Internet Engineering Task Force (IETF) Inband Flow Analyzer (IFA) telemetry information.