Network Health

The Network Health window allows for checking network connectivity or executing data traffic tests across the network, to ensure utmost network performance.

image2019-4-4_15-58-35.png

Connectivity tests send ping packets over the in-band network and report on connectivity issues between hosts, if found.

Connectivity tests can be run between all hosts in the fabric (using the All to All option), or between a group of hosts, each of which was randomly chosen from every switch (using the Random Candidate option).

Examples:

  • When connectivity is successful - “No Connectivity Issues Found” result will appear:

    image2019-4-4_16-1-19.png

  • When connectivity is unsuccessful - interfaces between which connectivity issues were found will be listed:

    image2019-4-4_16-2-38.png

Performance tests check data traffic across the network, in order to ensure utmost network performance.

There are seven types of tests that can be run individually or together, corresponding to the checkboxes checked. Description of each test is available upon clicking the question mark (“?”) of each test name.

Performance tests can be run between all hosts in the fabric (using the All to All option), or between a group of hosts, each of which was randomly chosen from every switch (using the Random Candidate option).

Examples:

  • The example below is of the RoCE bandwidth test run with “All to All” option:

    image2019-4-4_16-5-18.png

    • The X axis and the dots represent the hosts.

    • The Y axis represents the hosts’ test results. These results are the overall (aggregated) value of the RoCE bandwidth test results.

    • The gray average line represents the average result of all hosts’ test results.

  • The example below is of the RoCE latency test run with “All to All” option:

    image2019-4-4_16-10-2.png

  • The X axis and the dots represent the hosts.

  • The Y axis represents the hosts’ test results. These results are the average value of the RoCE latency test results.

  • The gray average line represents the average result of all hosts’ test results.

To run performance tests with more than one message size, use the “Performance” button under “Telemetry”. For more information, refer to “Performance”.

Warning

When running performance on Multicast, you should make the following settings in the host:

  1. Configuration of routing table: Run the following command on each host, for each interface: "route add -net 224.0.0.0/4 <interface>"

  2. Verify routing table configuration: Run the following command on each host, for each interface, to verify that it is registered for the multicast group: "route".

  3. You should see the following line for each interface in all hosts: "224.0.0.0 0.0.0.0 240.0.0.0 U 0 0 0 <interface>"

© Copyright 2023, NVIDIA. Last updated on Nov 16, 2023.