Triton provides Prometheus metrics indicating GPU and request statistics. By default, these metrics are available at http://localhost:8002/metrics. The metrics are only available by accessing the endpoint, and are not pushed or published to any remote server. The metric format is plain text so you can view them directly, for example:
$ curl localhost:8002/metrics
The tritonserver --allow-metrics=false option can be used to disable all metric reporting and --allow-gpu-metrics=false can be used to disable just the GPU Utilization and GPU Memory metrics. The --metrics-port option can be used to select a different port.
The following table describes the available metrics.
| Category | Metric | Description | Granularity | Frequency |
|---|---|---|---|---|
| GPU Utilization | Power Usage | GPU instantaneous power | Per GPU | Per second |
| Power Limit | Maximum GPU power limit | Per GPU | Per second | |
| Energy Consumption | GPU energy consumption in joules since Triton started | Per GPU | Per second | |
| GPU Utilization | GPU utilization rate (0.0 - 1.0) | Per GPU | Per second | |
| GPU Memory | GPU Total Memory | Total GPU memory, in bytes | Per GPU | Per second |
| GPU Used Memory | Used GPU memory, in bytes | Per GPU | Per second | |
| Count | Request Count | Number of inference requests received by Triton (each request is counted as 1, even if the request contains a batch) | Per model | Per request |
| Inference Count | Number of inferences performed (a batch of "n" is counted as "n" inferences) | Per model | Per request | |
| Execution Count | Number of inference batch executions (see Count Metrics) | Per model | Per request | |
| Latency | Request Time | Cumulative end-to-end inference request handling time | Per model | Per request |
| Queue Time | Cumulative time requests spend waiting in the scheduling queue | Per model | Per request | |
| Compute Input Time | Cumulative time requests spend processing inference inputs (in the framework backend) | Per model | Per request | |
| Compute Time | Cumulative time requests spend executing the inference model (in the framework backend) | Per model | Per request | |
| Compute Output Time | Cumulative time requests spend processing inference outputs (in the framework backend) | Per model | Per request |
For models that do not support batching, Request Count, Inference Count and Execution Count will be equal, indicating that each inference request is executed separately.
For models that support batching, the count metrics can be interpreted to determine average batch size as Inference Count / Execution Count. The count metrics are illustrated by the following examples:
-
Client sends a single batch-1 inference request. Request Count = 1, Inference Count = 1, Execution Count = 1.
-
Client sends a single batch-8 inference request. Request Count = 1, Inference Count = 8, Execution Count = 1.
-
Client sends 2 requests: batch-1 and batch-8. Dynamic batcher is not enabled for the model. Request Count = 2, Inference Count = 9, Execution Count = 2.
-
Client sends 2 requests: batch-1 and batch-1. Dynamic batcher is enabled for the model and the 2 requests are dynamically batched by the server. Request Count = 2, Inference Count = 2, Execution Count = 1.
-
Client sends 2 requests: batch-1 and batch-8. Dynamic batcher is enabled for the model and the 2 requests are dynamically batched by the server. Request Count = 2, Inference Count = 9, Execution Count = 1.