Performance Audit
Decision Tree
Performance issue → Where is it slow?
├─ API response time → Check DB queries, N+1, missing indexes
├─ Page load → Check bundle size, images, network waterfall
├─ Memory growing → Check for leaks (event listeners, closures, unclosed resources)
├─ Build/CI slow → Check caching, parallelization, unnecessary steps
└─ Unknown → Profile first, then follow the data
Phases
Phase 1: Measure (baseline) → Phase 2: Identify (profile) → Phase 3: Fix → Phase 4: Verify (compare to baseline)
Key Metrics
| Metric |
Target |
Tool |
| API response time (p95) |
<200ms |
Load test, APM |
| Time to First Byte |
<100ms |
curl -w, Lighthouse |
| First Contentful Paint |
<1.5s |
Lighthouse, WebPageTest |
| Memory usage |
Stable over time |
top, htop, profiler |
| DB query time |
<50ms per query |
Query logs, EXPLAIN |
| Bundle size (JS) |
<200KB gzipped |
npx bundlesize, webpack-bundle-analyzer |
Profiling Commands
# Node.js
node --prof app.js # V8 profiler
clinic doctor -- node app.js # Auto-diagnose (npm i -g clinic)
# Python
python -m cProfile -s cumtime app.py
py-spy record -o profile.svg -- python app.py
# Database
EXPLAIN ANALYZE SELECT ...; # PostgreSQL
Common Bottlenecks
| Pattern |
Symptom |
Fix |
| N+1 queries |
Slow list pages, many DB calls |
Eager loading, JOIN, batch fetch |
| Missing index |
Slow queries on large tables |
Add index on WHERE/JOIN columns |
| Unbounded query |
Memory spike, timeout |
Add LIMIT, pagination |
| Synchronous I/O |
Blocked event loop |
async/await, worker threads |
| Large payload |
Slow API, high bandwidth |
Pagination, field selection, compression |
| No caching |
Repeated expensive operations |
Redis, in-memory cache, HTTP cache headers |
| Unoptimized images |
Slow page load |
WebP, lazy loading, srcset |
| Memory leak |
Growing memory, eventual crash |
Heap snapshots, check event listeners/closures |
Database Query Analysis
-- PostgreSQL: Find slow queries
SELECT query, mean_exec_time, calls
FROM pg_stat_statements ORDER BY mean_exec_time DESC LIMIT 10;
-- Find missing indexes
SELECT relname, seq_scan, idx_scan
FROM pg_stat_user_tables WHERE seq_scan > idx_scan ORDER BY seq_scan DESC;
Output Format
[IMPACT: HIGH|MEDIUM|LOW] Category - Finding
Measured: Current value
Target: Expected value
Fix: Specific optimization
1---2name: performance-audit3description: Performance profiling with phased workflow, bottleneck patterns, and diagnostic queries. Use when diagnosing slow responses, high memory usage, or optimizing application performance.4---56# Performance Audit78## Decision Tree910```11Performance issue → Where is it slow?12 ├─ API response time → Check DB queries, N+1, missing indexes13 ├─ Page load → Check bundle size, images, network waterfall14 ├─ Memory growing → Check for leaks (event listeners, closures, unclosed resources)15 ├─ Build/CI slow → Check caching, parallelization, unnecessary steps16 └─ Unknown → Profile first, then follow the data17```1819## Phases2021```22Phase 1: Measure (baseline) → Phase 2: Identify (profile) → Phase 3: Fix → Phase 4: Verify (compare to baseline)23```2425## Key Metrics2627| Metric | Target | Tool |28|--------|--------|------|29| API response time (p95) | <200ms | Load test, APM |30| Time to First Byte | <100ms | curl -w, Lighthouse |31| First Contentful Paint | <1.5s | Lighthouse, WebPageTest |32| Memory usage | Stable over time | `top`, `htop`, profiler |33| DB query time | <50ms per query | Query logs, EXPLAIN |34| Bundle size (JS) | <200KB gzipped | `npx bundlesize`, webpack-bundle-analyzer |3536## Profiling Commands3738```bash39# Node.js40node --prof app.js # V8 profiler41clinic doctor -- node app.js # Auto-diagnose (npm i -g clinic)4243# Python44python -m cProfile -s cumtime app.py45py-spy record -o profile.svg -- python app.py4647# Database48EXPLAIN ANALYZE SELECT ...; # PostgreSQL49```5051## Common Bottlenecks5253| Pattern | Symptom | Fix |54|---------|---------|-----|55| N+1 queries | Slow list pages, many DB calls | Eager loading, JOIN, batch fetch |56| Missing index | Slow queries on large tables | Add index on WHERE/JOIN columns |57| Unbounded query | Memory spike, timeout | Add LIMIT, pagination |58| Synchronous I/O | Blocked event loop | async/await, worker threads |59| Large payload | Slow API, high bandwidth | Pagination, field selection, compression |60| No caching | Repeated expensive operations | Redis, in-memory cache, HTTP cache headers |61| Unoptimized images | Slow page load | WebP, lazy loading, srcset |62| Memory leak | Growing memory, eventual crash | Heap snapshots, check event listeners/closures |6364## Database Query Analysis6566```sql67-- PostgreSQL: Find slow queries68SELECT query, mean_exec_time, calls69FROM pg_stat_statements ORDER BY mean_exec_time DESC LIMIT 10;7071-- Find missing indexes72SELECT relname, seq_scan, idx_scan73FROM pg_stat_user_tables WHERE seq_scan > idx_scan ORDER BY seq_scan DESC;74```7576## Output Format7778```79[IMPACT: HIGH|MEDIUM|LOW] Category - Finding80 Measured: Current value81 Target: Expected value82 Fix: Specific optimization83```
BigPapiCB