Capacity for Backend Performances
Capacity planning is the estimation of how much compute, memory, storage and network a service needs to meet demand with safe headroom. It uses load test results, traffic forecasts and per-instance throughput to predict when limits will be hit, so teams add resources before users notice slowdowns.
Stage: Scale · Part of the Backend Performance roadmap · Last reviewed 19 November 2026
What to learn in Capacity
- Estimate peak requests per second from daily active users
- Find a single instance's saturation point with a load test
- Calculate how many instances are needed including failure headroom
- Forecast storage and bandwidth growth from historical metrics
- Weigh the cost of overprovisioning against the risk of an outage
Capacity interview questions
How would you estimate the servers needed for a service?
Forecast peak requests per second, measure how many one instance handles at acceptable latency in a load test, then divide and add headroom, often keeping utilisation well below saturation. Include spare capacity so losing one node or zone does not cause an outage.
Why keep headroom instead of running servers at full utilisation?
Near saturation, queues build and latency rises sharply, so small traffic spikes cause large slowdowns. Headroom absorbs bursts, failed nodes and deploys, and gives autoscaling time to react. Planned spare capacity is cheaper than an incident.
Where Capacity fits in the Backend Performance roadmap
Capacity belongs to the Scale stage (advanced level). Plan for growth.
Other units in this stage
Back to the full Backend Performance roadmap