Remote direct memory access (rdma) multipath

Patent metadata

Field	Value
Publication number	US-2025286812-A1
Application number	US-202519204775-A
Country	US
Kind code	A1
Filing date	May 12, 2025
Priority date	Sep 1, 2022
Publication date	Sep 11, 2025
Grant date	—

Abstract

Technologies for spreading a burst of data across multiple network paths in remote direct memory access (RDMA) over converged Ethernet (ROCE) and InfiniBand are described. A RDMA adapter receives, from a requestor device over a local interface, a request to send data of a transport flow directed to a target device over a network interface, and one or more parameters being related to a multipath selection by the network controller. The RDMA adapter sends a first burst of data of the transport flow via a first network path to the target device. The RDMA adapter identifies, using the one or more parameters, a second network path to the target device. The RDMA adapter sends the second burst of data to the target device on the second network path.

First claim

1 . (canceled)

2 . A network controller comprising:
a local interface coupled to a requestor device, wherein the network controller is to:
receive, from the requestor device over the local interface, a work request to send data of a transport flow directed to a target device over a network interface;
receive, from the requestor device over the local interface, one or more parameters related to a multipath selection by the network controller, the one or more parameters being based on a difference from an average round-trip time (RTT) of available paths to the target device;
send a first burst of data of the transport flow via a first network path to the target device;
determine, using the one or more parameters, that a second burst of data of the transport flow is to be sent to the target device on a different network path than the first network path;
identify, using the one or more parameters, a second network path to the target device; and
send the second burst of data to the target device on the second network path.

3 . The network controller of claim 2 , wherein the one or more parameters comprises at least one of i) a number of bursts since a last route change, ii) random entropy, iii) a requirement of an input fence, or iv) a comparison of a weight value, corresponding to a current route over the first network path, and other weight values corresponding to other available network paths to the target device.

4 . The network controller of claim 2 , wherein the network controller is further to:
measure a first RTT of the first network path; measure a second RTT of the second network path; determine the average RTT; update a first weight value based on the first RTT and the average RTT; and update a second weight value based on the second RTT and the average RTT.

5 . The network controller of claim 2 , wherein the network controller is further to:
measure an RTT of the first network path; determine whether the RTT is less than the average RTT of available network paths to the target device; and increase a first weight value responsive to the RTT being less than the average RTT; or decrease the first weight value responsive to the RTT being greater than the average RTT.

6 . The network controller of claim 2 , wherein the network controller is further to:
measure an RTT of the second network path; determine whether the RTT is less than the average RTT of available network paths to the target device; and increase a second weight value responsive to the RTT being less than the average RTT; or decrease the second weight value responsive to the RTT being greater than the average RTT.

7 . The network controller of claim 2 , wherein, to determine that the second burst of data of the transport flow is to be sent on a different network path than the first network path, the network controller is further to determine that a first weight value is greater than a weight value of at least one available network path to the target device.

8 . The network controller of claim 2 , wherein the transport flow uses Remote Direct Memory Access (RDMA) over Converged Ethernet (ROCE).

9 . The network controller of claim 2 , wherein the transport flow uses InfiniBand.

10 . The network controller of claim 2 , wherein the network controller is further to:
determine a first congestion metric for the first network path; determine a second congestion metric for the second network path; update a first weight value based on the first congestion metric; and update a second weight value based on the second congestion metric.

11 . The network controller of claim 2 , wherein the network controller is further to:
assign a first queue pair (QP) and a second QP to a multipath context, wherein the first burst of data and the second burst of data are stored in the first QP; schedule, by a scheduler of the network controller, the second burst of data from the second QP; schedule, by the scheduler, a third burst of data from the second QP; assign, by the network controller using the multipath context, a third network routing identifier to the third burst of data after the scheduling the third burst of data, the third network routing identifier corresponds to a third network path between the requester device and the target device; and send the third burst of data to the target device via the third network path.

12 . A computing system comprising:
a requestor device; and a network controller coupled to the requestor device over an interface, wherein the network controller is to:
receive, from the requestor device over the interface, a work request to send data of a transport flow directed to a target device over a network;
receive, from the requestor device over the interface, one or more parameters related to a multipath selection by the network controller, the one or more parameters being based on a difference from an average round-trip time (RTT) of available paths to the target device;
send a first burst of data of the transport flow via a first network path to the target device;
determine, using the one or more parameters, that a second burst of data of the transport flow is to be sent to the target device on a different network path than the first network path;
identify, using the one or more parameters, a second network path to the target device; and
send the second burst of data to the target device on the second network path.

13 . The computing system of claim 12 , wherein the transport flow uses Remote Direct Memory Access (RDMA) over Converged Ethernet (ROCE).

14 . The computing system of claim 12 , wherein the transport flow uses InfiniBand.

15 . The computing system of claim 12 , wherein to determine that the second burst of data of the transport flow is to be sent on a different network path than the first network path, the network controller is to determine that a first weight value is greater than a weight value of at least one available network path to the target device.

16 . The computing system of claim 12 , wherein the one or more parameters comprises at least one of i) a number of bursts since a last route change, ii) random entropy, iii) a requirement of an input fence, or iv) a comparison of a weight value, corresponding to a current route over the first network path, and other weight values corresponding to other available network paths to the target device.

17 . The computing system of claim 12 , wherein the network controller is further to:
measure a first RTT of the first network path; measure a second RTT of the second network path; determine the average RTT; update a first weight value based on the first RTT and the average RTT; and update a second weight value based on the second RTT and the average RTT.

18 . A network controller comprising:
a network adapter coupled to a requestor device over a local interface, wherein the network adapter is to:
receive, from the requestor device over the local interface, a request to send data of a transport flow directed to a target device over a network interface;
receive, from the requestor device over the local interface, one or more parameters related to a multipath selection by the network controller, the one or more parameters being based on a difference from an average round-trip time (RTT) of available paths to the target device;
send a first burst of data of the transport flow via a first network path to the target device;
identify, using the one or more parameters, a second network path to the target device; and
send a second burst of da

Assignees

• Mellanox Technologies Ltd

Inventors

• Friedman Yamin
• Borshteen Idan
• Moyal Roee
• Shpigelman Yuval

Classifications

• H04L45/24

  Multipath · CPC title

• H04L47/52

  by attributing bandwidth to queues · CPC title

• H04L43/0864

  Round trip delays · CPC title

• H04L41/0893

  Assignment of logical groups to network elements · CPC title

• H04L43/0876

  Network utilisation, e.g. volume of load or congestion level · CPC title