expose "Normalized Time Per Output Token" (NTPOT) metric

[kaushikmitr]

This pull request introduces a new metric to record the normalized time per output token (NTPOT) for inference models. The changes span across multiple files to implement this new metric and include updates to the metrics recording, registration, and testing.

New Metric Implementation:

• pkg/epp/metrics/metrics.go: Added a new histogram metric NormalizedTimePerOutputToken to record the normalized time per output token for each model and target model.
• pkg/epp/metrics/metrics.go: Implemented the RecordNormalizedTimePerOutputToken function to calculate and record the normalized time per output token.
• pkg/epp/metrics/metrics.go: Registered the new NormalizedTimePerOutputToken metric.

Integration with Existing Code:

• pkg/epp/handlers/server.go and pkg/epp/handlers/streamingserver.go: Added calls to RecordNormalizedTimePerOutputToken to record the metric when processing requests.

Testing:

• pkg/epp/metrics/metrics_test.go: Added test cases for RecordNormalizedTimePerOutputToken to verify the correct recording of the metric under various scenarios.
• pkg/epp/metrics/testdata/normalized_time_per_output_token_seconds_metric: Added test data for the new metric.

Documentation:

• site-src/guides/metrics.md: Updated the metrics documentation to include the new ntpot_seconds metric.

Tested by creating the following visualization

Added NTPO Inference Gateway Metric Dashboard

[k8s-ci-robot]

Hi @kaushikmitr. Thanks for your PR.

I'm waiting for a kubernetes-sigs member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

[netlify]

✅ Deploy Preview for gateway-api-inference-extension ready!

Name	Link
🔨 Latest commit	ba36b6e
🔍 Latest deploy log	https://app.netlify.com/sites/gateway-api-inference-extension/deploys/67eef23a6791b70008188919
😎 Deploy Preview	https://deploy-preview-643--gateway-api-inference-extension.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[JeffLuoo]

/ok-to-test

[liu-cong]

Suggested change

	log.FromContext(ctx).V(logutil.DEFAULT).Error(nil, "Request latency values are invalid for NTPOT calculation",
	log.FromContext(ctx).Error(nil, "Request latency values are invalid for NTPOT calculation",

[liu-cong]

Suggested change

	log.FromContext(ctx).V(logutil.DEFAULT).Error(nil, "Output token count must be positive for NTPOT calculation",
	log.FromContext(ctx).Error(nil, "Output token count must be positive for NTPOT calculation",

[liu-cong]

Can you add this to server.go as well?

[liu-cong]

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: kaushikmitr, liu-cong
Once this PR has been reviewed and has the lgtm label, please assign arangogutierrez for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[kaushikmitr]

/retest

[liu-cong]

/lgtm

[JeffLuoo]

/lgtm

[ahg-g]

Regarding the metrics name, can we align it with vllm naming convention: https://github.com/vllm-project/vllm/blob/main/docs/source/design/v1/metrics.md?plain=1#L37

normalized_time_per_output_token

[k8s-ci-robot]

New changes are detected. LGTM label has been removed.

[kaushikmitr]

updated metric name from ntpot to normalized time per output token