performance-optimization

Senior performance engineer who identifies bottlenecks and applies targeted optimizations with measurable impact.

Use this skill when

• Diagnosing or improving Core Web Vitals (LCP, INP, CLS)
• Reducing JavaScript bundle size or optimizing code splitting
• Implementing caching at any layer (HTTP, CDN, Redis, application)
• Optimizing database queries or fixing N+1 problems
• Debugging memory leaks in Node.js or browser
• Designing async processing pipelines for throughput

Core Web Vitals

LCP (Largest Contentful Paint) (target: < 2.5s)

LCP measures when the largest visible element finishes rendering. Usually a hero image, video, or large text block.

Common fixes:

1. Preload the LCP resource: <link rel="preload" as="image" href="/hero.webp">
2. Inline critical CSS, defer the rest. Use a critical CSS inliner such as beasties (the maintained fork of the archived critters).
3. Set fetchpriority="high" on the LCP image. Remove loading="lazy" from above-the-fold images.
4. Serve from CDN. Eliminate redirect chains.
5. Use 103 Early Hints to let the browser start fetching before HTML arrives.

<!-- Optimal LCP image -->
<img
  src="/hero.webp"
  srcset="/hero-480.webp 480w, /hero-800.webp 800w, /hero-1200.webp 1200w"
  sizes="(max-width: 600px) 480px, (max-width: 1024px) 800px, 1200px"
  width="1200"
  height="630"
  alt="Hero image"
  fetchpriority="high"
  decoding="async"
/>

INP (Interaction to Next Paint) (target: < 200ms)

INP replaced FID in March 2024. It measures the worst-case delay between user input and visual update across the entire page lifecycle.

Common fixes:

1. Break long tasks (>50ms) with scheduler.yield(). Fall back to a setTimeout promise where it is missing.
2. Move heavy computation to Web Workers.
3. Debounce rapid-fire events (scroll, resize, input). Use a 100-150ms debounce.
4. Avoid layout thrashing: batch DOM reads, then batch DOM writes.
5. Use content-visibility: auto on offscreen sections.

// Break a long task into yielding chunks
async function processItems(items: Item[]) {
  for (let i = 0; i < items.length; i++) {
    processItem(items[i]);
    if (i % 50 === 0) {
      // Yield to main thread every 50 items
      if ("scheduler" in globalThis && "yield" in scheduler) {
        await scheduler.yield();
      } else {
        await new Promise((resolve) => setTimeout(resolve, 0));
      }
    }
  }
}

CLS (Cumulative Layout Shift) (target: < 0.1)

Common fixes:

1. Always set explicit width and height on images and videos (browser calculates aspect ratio).
2. Reserve space for dynamic content: min-height on containers that load async.
3. Never inject content above existing content unless triggered by user interaction.
4. Use font-display: optional or preload fonts to prevent FOIT/FOUT shifts.
5. Use CSS aspect-ratio for responsive embeds: aspect-ratio: 16 / 9.

Bundle Analysis and Code Splitting

Analyze First

# Next.js: needs @next/bundle-analyzer wired into next.config
ANALYZE=true next build

# Webpack
npx webpack-bundle-analyzer dist/stats.json

# Vite
npx vite-bundle-visualizer

Look for: duplicate dependencies, moment.js (replace with dayjs), lodash (use lodash-es or per-function imports), large polyfills.

Code Splitting Patterns

// Route-based splitting (React)
const Dashboard = lazy(() => import("./pages/Dashboard"));
const Settings = lazy(() => import("./pages/Settings"));

// Component-level splitting for heavy UI
const HeavyChart = lazy(() => import("./components/HeavyChart"));

function App() {
  return (
    <Suspense fallback={<Skeleton />}>
      <Routes>
        <Route path="/dashboard" element={<Dashboard />} />
        <Route path="/settings" element={<Settings />} />
      </Routes>
    </Suspense>
  );
}

// Prefetch on hover for perceived instant loading
function NavLink({ to, children }: { to: string; children: React.ReactNode }) {
  const prefetch = () => {
    if (to === "/dashboard") import("./pages/Dashboard");
  };
  return <Link to={to} onMouseEnter={prefetch}>{children}</Link>;
}

Tree Shaking Essentials

• Use ESM (import/export), not CJS (require). CJS is not tree-shakeable.
• Set "sideEffects": false in package.json (or list files with side effects).
• Avoid barrel files (index.ts re-exporting everything): they defeat tree shaking.

Image Optimization

Format Selection

Next.js Image Component

import Image from "next/image";

// Automatically serves AVIF > WebP > JPEG, responsive sizes, lazy loaded
<Image
  src="/product.jpg"
  width={800}
  height={600}
  alt="Product photo"
  sizes="(max-width: 768px) 100vw, 50vw"
  priority={isAboveFold} // Sets fetchpriority="high", disables lazy load. Next.js 16 deprecates it in favor of preload.
/>

Sharp for Server-Side Processing

import sharp from "sharp";

await sharp(inputBuffer)
  .resize(1200, 630, { fit: "cover", position: "attention" }) // Smart crop
  .avif({ quality: 50 }) // AVIF at quality 50 ≈ JPEG at quality 80
  .toFile("output.avif");

Caching Strategies

HTTP Cache-Control Headers

# Static assets (hashed filenames): cache forever
Cache-Control: public, max-age=31536000, immutable

# API responses: revalidate after 60s, serve stale while revalidating
Cache-Control: public, max-age=60, stale-while-revalidate=300

# User-specific data: no shared cache
Cache-Control: private, max-age=0, must-revalidate

# Never cache
Cache-Control: no-store

Redis Caching Patterns

async function getCachedOrFetch<T>(
  key: string,
  fetcher: () => Promise<T>,
  ttlSeconds: number
): Promise<T> {
  const cached = await redis.get(key);
  if (cached) return JSON.parse(cached) as T;

  const data = await fetcher();
  // EX sets a TTL so stale entries expire. To stop a thundering herd on a
  // cold key, also take a short lock (SET lock:key 1 EX 10 NX) around fetcher().
  await redis.set(key, JSON.stringify(data), "EX", ttlSeconds);
  return data;
}

// Cache-aside with stale-while-revalidate pattern
async function getWithSWR<T>(key: string, fetcher: () => Promise<T>, ttl: number): Promise<T> {
  const cached = await redis.get(key);
  if (cached) {
    const parsed = JSON.parse(cached) as { data: T; fetchedAt: number };
    const age = Date.now() - parsed.fetchedAt;
    if (age > ttl * 500) {
      // Over 50% of TTL: revalidate in background. Catch errors so a failed
      // refresh does not crash the process with an unhandled rejection.
      fetcher()
        .then((fresh) =>
          redis.set(key, JSON.stringify({ data: fresh, fetchedAt: Date.now() }), "EX", ttl)
        )
        .catch((error) => logger.warn({ key, error }, "cache revalidation failed"));
    }
    return parsed.data;
  }
  const data = await fetcher();
  await redis.set(key, JSON.stringify({ data, fetchedAt: Date.now() }), "EX", ttl);
  return data;
}

Database Query Optimization

Index Strategy

1. Run EXPLAIN ANALYZE on every slow query. Look for Seq Scan on large tables.
2. Index columns used in WHERE, JOIN, ORDER BY. Composite indexes: leftmost prefix matters.
3. Covering indexes: include all SELECT columns so the DB reads only the index, not the table.

-- Composite index: supports WHERE user_id = ? AND created_at > ?
CREATE INDEX idx_orders_user_created ON orders (user_id, created_at DESC);

-- Covering index: query satisfied entirely from index
CREATE INDEX idx_orders_covering ON orders (user_id, created_at DESC)
  INCLUDE (total_amount, status);

-- Partial index: only index rows that matter
CREATE INDEX idx_orders_pending ON orders (created_at)
  WHERE status = 'pending';

N+1 Detection and Fix

// BAD: N+1: 1 query for users + N queries for orders
const users = await db.query("SELECT * FROM users LIMIT 100");
for (const user of users) {
  user.orders = await db.query("SELECT * FROM orders WHERE user_id = $1", [user.id]);
}

// GOOD: 2 queries total with a join or IN clause
const users = await db.query("SELECT * FROM users LIMIT 100");
const userIds = users.map((u) => u.id);
const orders = await db.query("SELECT * FROM orders WHERE user_id = ANY($1)", [userIds]);
const ordersByUser = Map.groupBy(orders, (o) => o.user_id);
for (const user of users) {
  user.orders = ordersByUser.get(user.id) ?? [];
}

Memory Leak Detection

Node.js

# Start with inspector
node --inspect --heapsnapshot-signal=SIGUSR2 dist/server.js

# Generate heap snapshot programmatically
kill -USR2 <pid>  # needs the flag above; Node then writes a .heapsnapshot file to cwd

Common Node.js leak sources:

1. Event listeners never removed. Use AbortController to clean up.
2. Closures capturing large objects. Null out references after use.
3. Global caches without eviction. Use lru-cache with max and ttl.
4. Unreferenced timers. Always clearInterval/clearTimeout on shutdown.

import { LRUCache } from "lru-cache";

// Bounded cache: max 1000 entries, 5-minute TTL
const cache = new LRUCache<string, object>({
  max: 1000,
  ttl: 1000 * 60 * 5,
});

Browser (Chrome DevTools)

1. Open Memory tab. Take a heap snapshot (Snapshot 1).
2. Perform the suspected leaking action (navigate, open modal, etc.).
3. Take Snapshot 2. Use "Comparison" view to see allocated-but-not-freed objects.
4. Look for Detached HTMLDivElement: DOM nodes removed from tree but still referenced in JS.
5. Filter by "Objects allocated between Snapshot 1 and 2" to find the leak source.

Connection Pooling

// Go database/sql: pool is built in
db, err := sql.Open("postgres", connStr)
db.SetMaxOpenConns(25)           // Match your DB's max_connections / number_of_instances
db.SetMaxIdleConns(10)           // Keep idle connections warm
db.SetConnMaxLifetime(5 * time.Minute)  // Rotate connections to rebalance after DB failover
db.SetConnMaxIdleTime(1 * time.Minute)  // Close idle connections to free DB resources

// Node.js with pg pool
import { Pool } from "pg";

const pool = new Pool({
  max: 20,                    // Max connections in pool
  idleTimeoutMillis: 30000,   // Close idle connections after 30s
  connectionTimeoutMillis: 5000, // Fail fast if no connection available
});

// Always release connections: use pool.query() for auto-release
const result = await pool.query("SELECT * FROM users WHERE id = $1", [userId]);

Async Processing Patterns

When an API handler does work that isn't needed for the response, move it out of the request path.

// BAD: user waits for email sending + analytics
app.post("/signup", async (req, res) => {
  const user = await createUser(req.body);
  await sendWelcomeEmail(user);      // 500ms
  await trackSignupEvent(user);      // 200ms
  res.json(user);                    // Total: 700ms+ for user
});

// GOOD: respond immediately, process async via queue
app.post("/signup", async (req, res) => {
  const user = await createUser(req.body);
  await queue.add("send-welcome-email", { userId: user.id });
  await queue.add("track-signup", { userId: user.id });
  res.json(user);                    // Total: ~50ms for user
});

Use BullMQ (Node.js), Celery (Python), or a message broker (RabbitMQ, SQS) for production queues. Always make queue consumers idempotent, because jobs may be retried.

Game Performance

Frame Budgets

At 60 FPS (16.6ms), budget roughly 8ms CPU (game logic, physics, animation, audio) and 8ms GPU (geometry, lighting, post-process, UI).

CPU Optimization

• Algorithmic: Spatial hashing for neighbor queries, early-out conditions, reduce O(n^2) to O(n log n).
• Cache-friendly data: Data-oriented design (Struct of Arrays over Array of Structs), process data linearly, minimize cache misses.
• Allocation: Object pooling, pre-allocate collections, avoid GC in hot paths.
• Threading: Offload to job systems, async loading, parallel processing.

GPU Optimization

• Draw calls: Static/dynamic batching, GPU instancing, merge meshes. Target <2000 on PC, <200 on mobile.
• Overdraw: Front-to-back rendering, occlusion culling, reduce transparency.
• Shaders: Reduce instruction count, use half precision, minimize texture samples.
• Geometry: LOD systems, mesh simplification, frustum culling.

Platform Targets

Mobile constraints:

• Thermal throttling and battery drain are primary limits
• Memory: 500MB-2GB. Draw calls: 100-200. Triangles: 100K-500K/frame. Texture memory: 200-500MB.
• Target 30-60 FPS stable.

VR requirements:

• Maintain 90 FPS constantly (dropped frames cause nausea).
• Single-pass stereo rendering, fixed foveated rendering, aggressive LOD, minimal post-processing.

Frame Rate Troubleshooting

1. Frame drops: Profile CPU vs GPU bound. Check for GC spikes. Move logic to FixedUpdate or coroutines. Add object pooling and LOD.
2. Long loads: Async loading with progress bar. Stream assets in background. Compress aggressively. Pre-warm caches.
3. Inconsistent pacing: Enable VSync. Use fixed timestep for physics. Spread heavy work across frames.

Profiling Tools