Capacity planning

📘What it really is

• Capacity planning makes sure you have enough resources before demand arrives, without paying for far more than needed.
• Inputs:
• current usage from metrics (Module 72): requests, CPU, memory, database size and connections, queue throughput;
• known limits from load and stress tests (Module 95): "one API copy handles about 150 requests per second at p95 under 800 ms; the database saturates at about 900 writes per second";
• growth forecast from the business: new cities, marketing campaigns, exam seasons, school holidays;
• lead time: how long it takes to add capacity (minutes for cloud servers, weeks for a bigger database plan or a provider contract).

• Headroom: run at a comfortable level (for example, peaks under 60 to 70% of limits), so spikes and failures (losing one of three copies) don't cause outages.

• Plan for known peaks: for the booking platform, board exam months and the start of school terms; run load tests at the expected peak beforehand, and pre-scale.

• Watch the growth trends of slow-growing resources: disk space, database size, table sizes, log volume, costs; alert well before limits.

• Limits outside your servers: provider rate limits (SMS, payment, AI APIs), connection limits, quotas on cloud accounts; request increases in advance.

• Write a simple capacity model: a spreadsheet linking business numbers (active students, bookings per day) to resource needs and costs (Module 95).