Performance & Profiling

Master an evidence-driven workflow to speed up Go programs without guesswork.

• Benchmarking & Micro-Optimizations:

  • Write realistic testing.B benchmarks and compare allocations with -benchmem.

  • Use table benchmarks to evaluate different algorithms (e.g., slice vs. map lookups).

  • Reduce GC pressure by reusing buffers (bytes.Buffer, sync.Pool) and avoiding unnecessary allocations.

• pprof & Tracing:

  • Capture CPU and heap profiles with net/http/pprof or on-demand files using pprof.StartCPUProfile.

  • Interpret flame graphs to spot hot paths, lock contention, and excessive allocations.

  • Use runtime/trace for end-to-end latency analysis and goroutine scheduling insights.

• Contended Resources:

  • Identify lock contention (-blockprofile, mutex profiles).

  • Replace coarse-grained locks with sharded maps, atomic operations, or message passing via channels.

• I/O & Serialization:

  • Batch I/O, use buffered readers/writers, and pre-size slices/maps.

  • Measure JSON vs. protobuf trade-offs; apply json.Encoder/Decoder with UseNumber to reduce conversions.

• Practical Lab:

  • Start with a slow service (JSON API + DB).

  • Baseline with benchmarks, add pprof endpoints, optimize hotspots, and demonstrate before/after metrics.