Congestion control for low latency datacenter networks
Systems and methods for controlling congestion of a data network are provided. An engine round-trip time (RTT) and a fabric RTT for a network flow are determined. An engine-based congestion window size for the flow is determined based on the engine RTT and a target engine RTT. A fabric-based congestion window size for the flow is determined based on the fabric RTT and a target fabric RTT. The smaller of the engine-based congestion window size and the fabric-based window size is selected for use in transmitting a future packet associated with the flow. The target engine RTT is determined based in part on the current congestion window used to transmit packets for the flow and/or the target fabric RTT is determined based on a number of hops packets associated with the flow traverse from a source to a destination associated with the flow.
1. A method comprising:
obtaining a total round-trip time (RTT) for a network flow;
determining an engine RTT for the network flow, and determining a fabric RTT for the network flow by subtracting the engine RTT from the total RTT;
determining an engine-based congestion window size for the flow based on the engine RTT and a target engine RTT;
determining a fabric-based congestion window size for the flow based on the fabric RTT and a target fabric RTT; and
selecting a smaller of the engine-based congestion window size and the fabric-based window size for use in transmitting at least one future packet associated with the flow, wherein the method further comprises at least one of:
determining the target engine RTT based in part on a current congestion window used to transmit packets for the flow; or
determining the target fabric RTT based on a number of hops packets associated with the flow traverse from a source network device for the flow to a destination network device associated with the flow.
2. The method of claim 1 , further comprising determining the number of hops traversed by packets associated with the flow.
3. The method of claim 1 , wherein the target fabric RTT is determined based on a number of hops packets associated with the flow traverse from a source network device for the flow and a destination network device associated with the flow by adding an amount of time to a base target fabric RTT equal to a time constant multiplied by the number of hops.
4. The method of claim 1 , wherein the target engine RTT is determined based in part on the current congestion window used to transmit packets for the flow by increasing a base target engine RTT by a value or factor that is inversely proportional to the current congestion window size for the flow.
5. The method of claim 1 , wherein the target engine RTT is determined based in part on the current congestion window used to transmit packets for the flow by increasing a base target engine RTT by a value or factor that is inversely proportional to a square root of the current congestion window size for the flow.
6. The method of claim 1 , wherein determining an engine-based congestion window size comprises decreasing the engine-based congestion window size in response to the engine RTT exceeding the target engine RTT.
7. The method of claim 1 , wherein determining the fabric-based congestion window size comprises decreasing the fabric-based congestion window size in response to the fabric RTT exceeding the target fabric RTT.
8. The method of claim 1 , wherein:
the target engine RTT is determined based in part on the current congestion window used to transmit packets for the flow; and
the target fabric RTT is determined based on a number of hops packets associated with the flow traverse from a source network device for the flow and a destination network device associated with the flow.
9. The method of claim 1 , further comprising updating the target fabric RTT for a flow in response to determining a change in the number of hops packets associated with the flow traverse.
10. The method of claim 1 , further comprising updating the target engine RTT in response to changing a current congestion window size for the flow.
11. A system comprising:
one or more processors; and
a non-transitory machine-readable medium comprising instructions stored therein, which when executed by the one or more processors, cause the one or more processors to perform operations comprising:
obtaining a total round-trip time (RTT) for a network flow;
determining an engine RTT for the network flow, and determining a fabric RTT for the network flow by subtracting the engine RTT from the total RTT;
determining an engine-based congestion window size for the flow based on the engine RTT and a target engine RTT;
determining a fabric-based congestion window size for the flow based on the fabric RTT and a target fabric RTT; and
selecting a smaller of the engine-based congestion window size and the fabric-based window size for use in transmitting at least one future packet associated with the flow, wherein the operations further comprise at least one of:
determining the target engine RTT based in part on a current congestion window used to transmit packets for the flow; or
determining the target fabric RTT based on a number of hops packets associated with the flow traverse from a source network device for the flow to a destination network device associated with the flow.
12. The system of claim 11 , wherein the operations further comprise determining the number of hops traversed by packets associated with the flow.
13. The system of claim 11 , wherein the target fabric RTT is determined based on a number of hops packets associated with the flow traverse from a source network device for the flow and a destination network device associated with the flow by adding an amount of time to a base target fabric RTT equal to a time constant multiplied by the number of hops.
14. The system of claim 11 , wherein the target engine RTT is determined based in part on the current congestion window used to transmit packets for the flow by increasing a base target engine RTT by a value or factor that is inversely proportional to the current congestion window size for the flow.
15. The system of claim 11 , wherein the target engine RTT is determined based in part on the current congestion window used to transmit packets for the flow by increasing a base target engine RTT by a value or factor that is inversely proportional to a square root of the current congestion window size for the flow.
16. The system of claim 11 , wherein determining an engine-based congestion window size comprises decreasing the engine-based congestion window size in response to the engine RTT exceeding the target engine RTT.
17. The system of claim 11 , wherein determining the fabric-based congestion window size comprises decreasing the fabric-based congestion window size in response to the fabric RTT exceeding the target fabric RTT.
18. The system of claim 11 , wherein:
the target engine RTT is determined based in part on the current congestion window used to transmit packets for the flow; and
the target fabric RTT is determined based on a number of hops packets associated with the flow traverse from a source network device for the flow and a destination network device associated with the flow.
19. The system of claim 11 , wherein the operations further comprise updating the target fabric RTT for a flow in response to determining a change in the number of hops packets associated with the flow traverse.
20. The system of claim 11 , wherein the operations further comprise updating the target engine RTT in response to changing a current congestion window size for the flow.