Troubleshoot AF Analysis Performance

PI System Explorer provides important service and analysis information for troubleshooting PI Analysis Service performance. The metrics are most useful when teams connect them to specific analyses, templates, assets, and dependencies.

Start with service-level indicators

Review available analysis-service metrics such as calculation lag, skipped calculations, service health, and workload behavior.

Use the metrics to determine whether the issue is broad or isolated.

Find the analyses behind the symptom

If lag or skipped calculations increase, identify which analyses or templates consume unusual runtime or fail repeatedly.

A service-level warning does not identify the business effect by itself.

Review template scale

A calculation that is acceptable on one asset can become expensive when deployed across thousands of elements.

Check:

• Trigger frequency

• Number of deployed analyses

• Input retrieval behavior

• Expression complexity

• Recalculation behavior

Review long dependency chains

An analysis can depend on outputs from other analyses. Delays upstream can propagate downstream.

Trace important chains and identify whether the observed lag starts earlier in the dependency path.

Check bad or missing inputs

Repeated input failures can create analysis errors and unnecessary troubleshooting.

Confirm data freshness and quality for key inputs before changing the expression.

Review recent changes

Performance problems often start after a template deployment, schedule change, expression modification, or new asset rollout.

Compare the start of the problem with configuration history.

Optimize based on evidence

Possible actions include reducing unnecessary trigger frequency, simplifying expensive logic, correcting bad inputs, distributing workload appropriately, or redesigning a calculation chain.

Do not optimize only because one metric is high. Confirm the actual bottleneck.

Verify downstream impact

Before disabling or modifying a slow analysis, identify the displays and other calculations that use its output.

The practical goal is not only a healthy Analysis Service. It is a calculation environment that meets required timing and remains understandable to the teams that support it.