Skip to main content
Version: Next

Check Metrics in Monitoring

Use NuFi's Monitoring menu to view real-time performance metrics of serving models.

Access the Monitoring Dashboard​

In the left sidebar, click Monitoring. In the Device Overview tab, select a GPU or RNGD node to view its resource usage.

Monitoring initial screen

Available Metrics​

The Monitoring dashboard provides the following metrics:

  • Request Count: number of requests
  • Error Rate: error rate
  • Latency / TTFT: response latency and time to first token
  • TPS (tokens/sec): token throughput per second
  • Accelerator Utilization: GPU/NPU utilization
  • Power (W): power consumption
  • Performance/Watt (TPS/W): throughput per watt

Monitoring — serving metrics

Tutorial Complete​

You have checked the serving response and cluster resource metrics.