Capacity Planning
1 min read
Size infrastructure for current load plus growth with headroom for spikes.
Size servers, memory, and storage from your QPS estimates. Add 2-3x headroom for growth and spikes.
How It Works
Capacity planning translates QPS and storage estimates into server counts and infrastructure decisions. CPU-bound services: estimate requests per core (100-1000 for API servers), divide peak QPS by that number, add 30% headroom. Memory-bound services (caches): total working set size / memory per instance. Always plan for 2-3x current peak.
Real-World Example
For a chat system handling 50M concurrent connections: each WebSocket connection uses ~10KB of memory. 50M x 10KB = 500GB. With 64GB per server, that is ~8 servers for connections alone — plus application logic, plan for 15-20 servers.
Test Yourself
Scenario: A live video streaming service expects 8M concurrent viewers at peak with an average bitrate of 5 Mbps. Each edge CDN server can push 40 Gbps. Size the edge fleet with headroom.
Get notified when we launch
One email when the full practice product is live. No spam.