software-engineering-eng-perf

What problem does it solve?

Performance investigations fail when teams rely on guesses instead of measurements, leading to wasted effort and persistent latency or throughput issues.

Core Features & Use Cases

• Systems performance diagnosis (USE Method): Evaluate utilization, saturation, and errors across CPU, memory, disk I/O, network, connections, and thread pools to determine the bottleneck.
• Latency and service diagnosis (RED Method): Analyze Rate, Errors, and Duration using p50/p90/p95/p99 to pinpoint slow endpoints and tail-latency problems.
• Profiling-driven root cause analysis: Use CPU profiling and flame graphs, plus memory and database investigation patterns (e.g., N+1, caching strategy, pool exhaustion) to produce actionable fixes.
• Verification via re-measurement and load testing: Validate improvements using the same metrics and a structured load-testing protocol (smoke/load/stress/soak/spike).

Example use cases: diagnosing an API p99 latency regression, finding the CPU hotspot behind a flame graph, isolating memory leaks over time, and optimizing database queries and caching behavior.

Quick Start

Ask the AI to diagnose your performance issue by requesting the current p99 latency, the suspected endpoint or operation, the SLO target, and the relevant measurements or traces.