Application Monitoring (APM)

Application Performance Monitoring (APM) provides deep visibility into application behavior by automatically instrumenting code to track request traces, database queries, external HTTP calls, and resource utilization. APM tools capture the full lifecycle of each request as a distributed trace, breaking it into spans that represent individual operations (route handler, database query, cache lookup, external API call) with precise timing information.

Modern APM solutions follow the OpenTelemetry (OTel) standard, which provides vendor-neutral instrumentation APIs, SDKs, and auto-instrumentation libraries. OTel collects three signals: traces (distributed request paths), metrics (counters, gauges, histograms for throughput, latency, error rates), and logs (correlated via trace context). This data is exported to backends like Datadog, New Relic, Grafana Tempo, Jaeger, or AWS X-Ray for visualization and alerting.

Key APM metrics include request throughput (requests per second), error rate (percentage of failed requests), and latency percentiles (p50, p95, p99). The RED method (Rate, Errors, Duration) and USE method (Utilization, Saturation, Errors) provide frameworks for monitoring services and resources respectively. APM enables identifying slow database queries, N+1 query patterns, external service latency, memory leaks, and connection pool exhaustion. Alerting should be based on SLOs (Service Level Objectives) such as 'p99 latency under 500ms for 99.9% of 5-minute windows.'