Backend Performance Profiling Techniques

Explain how to identify backend performance bottlenecks.

Profiling Process

1. Measure: Establish a baseline and confirm the problem exists.
2. Profile: Identify hot spots — which code paths or queries consume the most time.
3. Optimize: Address the biggest bottleneck first.
4. Verify: Re-measure to confirm the improvement.

Tools

APM (Application Performance Monitoring)

Datadog, New Relic, and Sentry Performance provide distributed tracing and automatic slow transaction detection.

Distributed Tracing

OpenTelemetry + Jaeger/Zipkin traces cross-service request chains to identify which service is the bottleneck.

Database Slow Query Logs

MySQL slow_query_log and PostgreSQL pg_stat_statements log queries that exceed a time threshold.

Flame Graph

Visually shows CPU time distribution across the call stack, quickly identifying hot functions.

Common Bottlenecks

N+1 queries, missing indexes, heavy synchronous I/O, memory leaks (GC pressure), lock contention.