performance-analysis_skill

Overview

This skill covers measurement approaches, profiling tools, optimization patterns, and capacity planning for diagnosing and improving performance. It teaches a measurement-first workflow to establish baselines, identify bottlenecks, implement targeted fixes, and validate results. Use it to create reproducible performance processes and capacity plans.

How this skill works

The skill inspects application, system, and infrastructure layers using appropriate profilers and metrics. It guides use of CPU, memory, and I/O profiling, applies bottleneck identification methods (USE, RED), and maps findings to optimization patterns and capacity planning steps. Emphasis is on measuring before changing, validating after fixes, and documenting outcomes.

When to use it

• Diagnosing slow responses, high CPU, or memory issues
• Establishing performance baselines before optimization
• Identifying bottlenecks across app, DB, or infra
• Planning capacity for expected growth or spikes
• Validating performance improvements after changes

Best practices

• Always measure first and record a baseline before changes
• Profile in production-like environments and keep overhead low
• Use latency percentiles (p95, p99) rather than averages
• Monitor continuously and correlate metrics across layers
• Set and enforce performance budgets in CI

Example use cases

• Find and fix a CPU-bound function using sampling profilers and flame graphs
• Locate a memory leak with heap snapshots and allocation tracking
• Resolve I/O bottlenecks by measuring p99 latency and optimizing queries or cache usage
• Plan capacity by ramp, stress, and soak tests to determine throughput and recovery time
• Implement quick wins like caching, compression, and connection pooling to reduce latency