APM Tips

Glossary

APM (application performance monitoring)

The practice and the tooling for measuring how a running application behaves for its users, chiefly response time, error rate and throughput, and tracing slow or failed work to its cause.

Updated

APM tools collect three kinds of signal from a running application: traces (a request and every call it made, with timings), metrics (counts and durations aggregated over time) and logs (events with context). Together they answer the operational questions: is the app up, is it fast enough, what is failing, and where is the time going.

The term predates the word observability and is often used for the commercial products in the space. Azure Monitor Application Insights is Microsoft’s APM service; OpenTelemetry is the vendor-neutral standard most APM products now accept as input.

A useful APM setup is judged by its percentiles rather than its averages: a p95 or p99 response time describes the slow requests users actually notice. See the request latency query for one way to measure them.

Where it comes up

  • Percentile (p50, p95, p99)

    The value a given share of measurements fall at or below; a p95 latency of 800 ms means 95% of requests took 800 ms or less, and one in twenty took longer.

  • Dependency

    A call an app makes to something outside it, such as an HTTP API, a database, a queue or a cache, recorded in the dependencies table with its target, duration and result.

  • Sampling

    Keeping a representative fraction of telemetry instead of all of it, to cut ingestion volume and cost while keeping counts and percentiles statistically useful.

All glossary terms