Networking Minerva User Guide
Networking Minerva User Guide
Breadcrumbs

Preconfigured Dashboards

NVIDIA Minerva includes preconfigured dashboards for monitoring AI factory infrastructure. Dashboards are organized into two suites: AI Factory Networking covers GPU, NVLink, and Spectrum-X Ethernet fabric health; Full Stack correlates infrastructure health with active job workloads.

  • AI Factory Networking: provides multi-layer visibility across the entire fabric:

    • Node and GPU: per-node and per-GPU performance tracking.

    • NVLink: monitoring of NVLink interconnect health between GPUs.

    • Spectrum-X: observability into the Spectrum-X Ethernet fabric, including RDMA performance, switch-localized congestion, and port-level statistics.

  • Full Stack: correlates infrastructure health with specific workload IDs. By mapping GPU utilization, NVLink errors, and thermal data directly to active jobs, you can determine whether a performance dip is caused by hardware degradation or inefficient job scheduling.

Import Dashboards

To import the dashboards:

  1. In the NVIDIA Minerva UI, click settings.png Settings in the upper right.

  2. Open the Actions menu and select Import.

  3. Click Select file and choose the JSON file for the dashboard bundle you want to import.

  4. Click Import.

Last updated: