NVIDIA Messaging Accelerator (VMA) Documentation Rev 9.8.40
NVIDIA Messaging Accelerator (VMA) Documentation Rev 9.8.40

Using sockperf with VMA

Sockperf is VMA's sample application for testing latency and throughput over a socket API. The precompiled sockperf binary is located in /usr/bin/sockperf.

For detailed instructions on how to optimally tune your machines for VMA performance, please see the Tuning Guide and VMA Performance Tuning Guide.

To run a sockperf UDP test:

  • To run the server, use:

    Copy
    Copied!
                

    LD_PRELOAD=libvma.so sockperf sr -i <server ip>

  • To run the client, use:

    Copy
    Copied!
                

    LD_PRELOAD=libvma.so sockperf <sockperf test> –i <server ip>

Where:

  • <server ip> is the IP address of the server

  • <sockperf test> is the test you want to run, for example, pp for the ping-pong test, tp for the throughput test, and so on. (Use sockperf -h to display a list of all available tests.)

To run a sockperf TCP test:

  • To run the server, use:

    Copy
    Copied!
                

    LD_PRELOAD=libvma.so sockperf sr -i <server ip> --tcp

  • To run the client, use:

    Copy
    Copied!
                

    LD_PRELOAD=libvma.so sockperf <sockperf test> –i <server ip> --tcp

For optimal performance, please refer to Basic Performance Tuning.

  1. Run sockperf server on Host A:

    Copy
    Copied!
                

    LD_PRELOAD=libvma.so sockperf sr

  2. Run sockperf client on Host B:

    Copy
    Copied!
                

    LD_PRELOAD=libvma.so sockperf pp –i <server_ip>

© Copyright 2023, NVIDIA. Last updated on Nov 3, 2023.