Optimizing Performance

Optimizing Performance

When to Load

• Trigger: Diagnosing slowness, profiling, caching strategies, reducing load times, bundle size optimization
• Skip: Correctness-focused work where performance is not a concern

Performance Optimization Workflow

Copy this checklist and track progress:

Performance Optimization Progress:
- [ ] Step 1: Measure baseline performance
- [ ] Step 2: Identify bottlenecks
- [ ] Step 3: Apply targeted optimizations
- [ ] Step 4: Measure again and compare
- [ ] Step 5: Repeat if targets not met

Critical Rule: Never optimize without data. Always profile before and after changes.

Step 1: Measure Baseline

Profiling Commands

# Node.js profiling
node --prof app.js
node --prof-process isolate*.log > profile.txt

# Python profiling
python -m cProfile -o profile.stats app.py
python -m pstats profile.stats

# Web performance
lighthouse https://example.com --output=json

Step 2: Identify Bottlenecks

Common Bottleneck Categories

| Category | Symptoms | Tools | | -------- | -------------------------------- | ------------------------------- | | CPU | High CPU usage, slow computation | Profiler, flame graphs | | Memory | High RAM, GC pauses, OOM | Heap snapshots, memory profiler | | I/O | Slow disk/network, waiting | strace, network inspector | | Database | Slow queries, lock contention | Query analyzer, EXPLAIN |

Step 3: Apply Optimizations

Frontend Optimizations

Bundle Size:

// ❌ Import entire library
import _ from "lodash";

// ✅ Import only needed functions
import debounce from "lodash/debounce";

// ✅ Use dynamic imports for code splitting
const HeavyComponent = lazy(() => import("./HeavyComponent"));

Rendering:

// ❌ Render on every parent update
function Child({ data }) {
  return ;
}

// ✅ Memoize when props don't change
const Child = memo(function Child({ data }) {
  return ;
});

// ✅ Use useMemo for expensive computations
const processed = useMemo(() => expensiveCalc(data), [data]);

Images:

Backend Optimizations

Database Queries:

-- ❌ N+1 Query Problem
SELECT * FROM users;
-- Then for each user:
SELECT * FROM orders WHERE user_id = ?;

-- ✅ Single query with JOIN
SELECT u.*, o.*
FROM users u
LEFT JOIN orders o ON u.id = o.user_id;

-- ✅ Or use pagination
SELECT * FROM users LIMIT 100 OFFSET 0;

Caching Strategy:

// Multi-layer caching
const getUser = async (id) => {
  // L1: In-memory cache (fastest)
  let user = memoryCache.get(`user:${id}`);
  if (user) return user;

  // L2: Redis cache (fast)
  user = await redis.get(`user:${id}`);
  if (user) {
    memoryCache.set(`user:${id}`, user, 60);
    return JSON.parse(user);
  }

  // L3: Database (slow)
  user = await db.users.findById(id);
  await redis.setex(`user:${id}`, 3600, JSON.stringify(user));
  memoryCache.set(`user:${id}`, user, 60);

  return user;
};

Async Processing:

// ❌ Blocking operation
app.post("/upload", async (req, res) => {
  await processVideo(req.file); // Takes 5 minutes
  res.send("Done");
});

// ✅ Queue for background processing
app.post("/upload", async (req, res) => {
  const jobId = await queue.add("processVideo", { file: req.file });
  res.send({ jobId, status: "processing" });
});

Algorithm Optimizations

// ❌ O(n²) - nested loops
function findDuplicates(arr) {
  const duplicates = [];
  for (let i = 0; i  4s    |
| INP    |  500ms |
| CLS    |  0.25  |
| TTFB   |  1.8s  |

### API Performance

| Metric      | Target  |
| ----------- | ------- |
| P50 Latency | < 100ms |
| P95 Latency | < 500ms |
| P99 Latency | < 1s    |
| Error Rate  | < 0.1%  |

## Validation

After optimization, validate results:

Performance Validation:

• [ ] Metrics improved from baseline
• [ ] No functionality regressions
• [ ] No new errors introduced
• [ ] Changes are sustainable (not one-time fixes)
• [ ] Performance gains documented

If targets not met, return to Step 2 and identify remaining bottlenecks.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [CloudAI-X](https://github.com/CloudAI-X)
- **Source:** [CloudAI-X/claude-workflow-v2](https://github.com/CloudAI-X/claude-workflow-v2)
- **License:** MIT

Install and usage instructions live in the source repository linked above.