Yandex Cloud
Documentation
PresentationContact UsTry it for free
© 2026 Direct Cursus Technology L.L.C.
Yandex Compute Cloud
    • All guides
      • Creating a VM with a GPU
      • Installing NVIDIA drivers
      • NVIDIA driver update guide
      • Enabling MIG on a VM with a GPU
      • Managing maintenance of GPU VMs
        • Creating a GPU cluster
        • Adding a VM to a GPU cluster
        • Updating a GPU cluster
        • Getting GPU cluster info
        • Configuring GPU cluster access permissions
        • Deleting a GPU cluster
        • Testing a GPU cluster physical state
        • Running parallel tasks in a GPU cluster
        • Testing InfiniBand throughput
    • Viewing operations with resources
    • Viewing metrics Monitoring
  • Yandex Container Solution
  • Access management
  • Pricing policy
  • Terraform reference
  • Monitoring metrics
  • Audit Trails events
  • Release notes

In this article:

  • Test the state of InfiniBand ports
  • Test network performance
  1. Step-by-step guides
  2. GPU
  3. GPU clusters
  4. Testing a GPU cluster physical state

Testing a GPU cluster physical state

Written by
Yandex Cloud
Updated at July 31, 2026
View in Markdown
  • Test the state of InfiniBand ports
  • Test network performance

Test the state of InfiniBand portsTest the state of InfiniBand ports

  1. Connect to the VM over SSH.

  2. Install the infiniband-diags package:

    sudo apt update
    sudo apt install infiniband-diags
    
  3. Run the ibstatus command:

    Result:

    state:           4: ACTIVE
    phys state:      5: LinkUp
    
  4. Make sure the phys state parameter is set to LinkUp for all ports.

Test network performanceTest network performance

To test the data transfer rate between GPUs on different VMs:

  1. Install the perftest package on each test VM:

    sudo apt install perftest
    
  2. Connect to the first VM over SSH.

  3. Run this command:

    ib_send_bw --report_gbits
    
  4. Connect to the second VM over SSH.

  5. Run this command:

    ib_send_bw <first_VM_internal_IP> --report_gbits
    

    Result:

    #bytes     #iterations    BW peak[Gb/sec]    BW average[Gb/sec]   MsgRate[Mpps]
    65536      1000             245.54             244.08             0.465536
    
  6. Make sure the output shows non-zero values for these parameters:

    • BW average[Gb/sec]: Average transfer rate
    • MsgRate[Mpps]: Message frequency

Useful linksUseful links

Questions about GPUs

Was the article helpful?

Previous
Deleting a GPU cluster
Next
Running parallel tasks in a GPU cluster
© 2026 Direct Cursus Technology L.L.C.