System design interview guide/Printable worksheet
System design interview practice sheet
A free, two-page PDF for a 45-minute practice round. It gives you a time plan, quick sizing formulas, five questions and a checklist. No signup is needed.
The 45-minute plan
- 5 min
Clarify the problem
Agree on users, requirements, quality targets and the assumptions that shape the design.
- 5 min
Estimate the load
Calculate average and peak requests, data growth and the read-to-write ratio.
- 10 min
Draw the main paths
Show services, stores, caches and queues. Trace one request and one write.
- 15 min
Choose one hard part
Compare two plausible designs for the hardest constraint and name the cost you accept.
- 5 min
Break and change it
Trace a crash, duplicate, timeout or spike. Then adapt to a changed requirement.
- 5 min
State the guarantees
Explain what the design promises, where it can fail and what you would improve next.
Quick sizing formulas
- Average requests per second = requests per day divided by 86,400.
- Peak requests per second = average requests per second multiplied by the peak factor.
- Raw storage per year = writes per day multiplied by bytes per record and 365.
For example, 10 million requests a day is about 116 requests per second on average. At a five-times peak, plan for about 580 requests per second.
Estimate QPS, storage and bandwidth with the free calculator →Five questions to practise
- 1.
Design a URL shortener. Identifiers, read-heavy traffic, cache invalidation and analytics.
- 2.
Design an API rate limiter. Burst handling, shared counters and failure policy.
- 3.
Design a news feed. Fan-out, celebrity accounts, inactive readers and deletes.
- 4.
Design a payment system. Retries, idempotency and out-of-order provider events.
- 5.
Design a distributed job queue. Durable work, duplicate delivery, worker crashes and backlogs.
Work through each as an interview, then compare your decisions with a complete explanation. The exercises are free and work without an account.
Published by SysGeeks. The sheet supports practice; it does not predict how a particular interviewer will score an answer.