Performance Optimizing

Optimization Workflow

1. Measure first - Don't optimize without data
2. Find bottlenecks - Profile to identify hot paths
3. Fix the biggest issue - Pareto principle (80/20)
4. Measure again - Verify improvement
5. Document - Record what changed and why

Common Bottlenecks

Database

| Problem | Solution | | ---------------- | --------------------------------- | | N+1 queries | Eager loading, batching | | Missing indexes | Add indexes on WHERE/JOIN columns | | Full table scans | Optimize queries, add indexes | | Over-fetching | Select only needed columns |

// Bad - N+1 queries
for (const user of users) {
  const posts = await db.posts.findByUserId(user.id);
}

// Good - single query with join
const usersWithPosts = await db.users.findAll({
  include: ['posts'],
});

Memory

| Problem | Solution | | --------------- | ----------------------------- | | Memory leaks | Clear references, use WeakMap | | Large objects | Stream processing, pagination | | Duplicated data | Normalize, use references |

CPU

| Problem | Solution | | ---------------------- | ---------------------- | | Blocking operations | Use async/workers | | Redundant computation | Memoization, caching | | Inefficient algorithms | Better data structures |

// Bad - O(n) lookup on every iteration
for (const item of items) {
  if (list.includes(item.id)) { ... }
}

// Good - O(1) lookup with Set
const idSet = new Set(list);
for (const item of items) {
  if (idSet.has(item.id)) { ... }
}

Network

| Problem | Solution | | ----------------- | ----------------------- | | Too many requests | Batching, HTTP/2 | | Large payloads | Compression, pagination | | No caching | HTTP caching, CDN |

Quick Wins

1. Add indexes on frequently queried columns
2. Enable compression (gzip/brotli)
3. Implement caching at appropriate layers
4. Use pagination for large datasets
5. Lazy load non-critical resources

Anti-Patterns

• Premature optimization without measurement
• Micro-optimizations that hurt readability
• Caching everything (cache invalidation is hard)
• Optimizing cold paths