date: 2025-07-23
topics: [nodejs, performance, clustering, event-loop, backend]
day: 7
📘 Day 7 – Node.js Cluster & Performance Benchmarking
✅ What I Worked On
- Built a benchmarking setup using Node.js + Cluster + Worker Threads
- Benchmarked sync vs async route performance using
autocannon
- Integrated BullMQ for offloading heavy tasks to background workers
- Used
os.cpus() to scale server instances across CPU cores
📚 What I Learned
| Scenario | Requests/sec | Timeouts | Avg Latency |
|---|
| Async Route | ✅ ~6.2k | ✅ ~80 | ✅ ~125ms |
| Sync Route (Blocking) | ❌ ~240 | ❌ 3k+ | ❌ up to 10s |
- 🧠 Async I/O in Node is highly performant when you avoid blocking operations
- ⚠️ Sync code kills event loop → adds timeouts and high latency
- ✅ Use cluster module for horizontal CPU scaling
- ✅ Offload CPU-intensive work using:
worker_threads- Background queues (BullMQ, Redis)
❌ Blockers
- Initial trouble spawning isolated workers without shared memory leak
- Needed to fine-tune autocannon settings for accurate measurements
🧠 Reflection
This wasn't just coding — this was understanding Node.js under the hood. Benchmarking clarified why certain design patterns (queues, cluster, async APIs) aren't just optional — they’re critical for real-world backend performance.
🔗 References / Code