Site Reliability Engineer interview question
How do you approach capacity planning?
Forecast demand from historical growth and planned events, measure resource per unit of load, model peak and failure headroom, identify the first saturating dependency, load-test assumptions, and set scaling lead times.
Direct answer
What a strong answer covers
How it works
Capacity planning turns a demand forecast into resource decisions with lead time. Demand comes from historical growth, seasonality, and planned events such as launches or campaigns. For each service, the resource consumed per unit of load is measured, so a forecast of requests per second becomes CPU, memory, connections, and storage. The plan models peak with headroom for a zone loss or a failed deploy, identifies which dependency saturates first, and validates the assumptions with a load test at the projected level. Scaling lead times, from autoscaler minutes to hardware or quota weeks, decide how far ahead each decision must be made.
Example
Ahead of a holiday campaign forecast at 3x normal traffic, the team measured that each 1,000 requests per second cost 6 cores and 120 database connections; the database connection limit would saturate at 2.2x. A load test confirmed it, so the fix was connection pooling through PgBouncer and a read replica, ordered three weeks ahead because the cloud quota increase took a week to approve.
Common mistakes
- Planning from average load and being surprised by the peak.
- Scaling the web tier while the database is the first thing to saturate.
- Discovering a cloud quota limit on launch day.
A useful answer structure
Answer from your own work.
Do not memorise a script: use only examples you can explain honestly in follow-up questions.
Open the desktop workspace and rehearse a clear answer in your own words.