GamifiedGamified challengesPracticeRaceLeaderboardPricing

Learn data structures by watching them move, then doing it yourself. Press M anytime to toggle sound.

A product of Amypo Technologies, Coimbatore.

Learn

  • All topics
  • Learning path
  • How it works
  • About dotprep

Practice

  • Practice arena
  • Daily challenge
  • Gamified challenges
  • Race mode
  • Leaderboard

Interview prep

  • Java interview
  • Coding questions
  • Company questions
  • System design
  • Jobs & careers

You

  • Dashboard
  • Settings
  • Pricing
  • Create account

© 2026 dotprep

Content last updated 20 November 2026

GamifiedGamified challengesPracticeRaceLeaderboardPricing
System designHow to approach a design interview

How to approach a design interview

A repeatable 45-minute framework: clarify requirements, estimate, sketch, deep-dive and discuss trade-offs.

beginner 10 min read 4 interview questions

On this page

1/5
  1. 01The framework
  2. 02Back-of-the-envelope estimation
  3. 03Common mistakes
  4. 04Interview questions
  5. 05Quick check
Back to top

The framework

A design interview has no single right answer. The interviewer is watching how you get to a design: whether you ask the right questions, make reasonable assumptions, and can explain the trade-offs you chose. A fixed structure keeps you from drawing boxes before you know what you're building.

StepTimeWhat you do
1. Requirements~5 minFunctional: what users can do. Non-functional: scale, latency, availability, consistency. Agree on what's out of scope.
2. Estimates~5 minUsers, requests per second, read/write ratio, storage per year, bandwidth. Round numbers are fine.
3. API and data model~5 minThe main endpoints and the entities with their key fields.
4. High-level design~10 minClients, load balancer, services, databases, caches, queues. Walk one request end to end.
5. Deep dive~15 minPick the hardest one or two parts (often the one the interviewer steers you to) and go into detail.
6. Wrap up~5 minBottlenecks, failure modes, monitoring, and what you'd do with more time.

Say your assumptions out loud and write them down. "Let's assume 100 million daily users and a 100:1 read/write ratio" turns a vague prompt into a problem you can size.

Back-of-the-envelope estimation

Estimates decide the design: 100 writes per second fits on one database, 100,000 does not. Keep a few numbers in your head and round aggressively.

  • A day has about 86,400 ≈ 10⁵ seconds, so 1 million requests per day ≈ 12 per second.
  • Peak traffic is often 2–3× the average.
  • 1 KB × 1 million = 1 GB. 1 KB × 1 billion = 1 TB.
  • One commodity server can usually handle thousands of simple requests per second; a single SQL primary handles thousands of writes per second for small rows.
text
Example: a photo app with 10M daily active users, each uploading 2 photos/day
writes = 10M × 2 / 10⁵ s ≈ 200 uploads/s (peak ≈ 600/s)
storage = 20M × 500 KB / day ≈ 10 TB/day ≈ 3.6 PB/year
reads = 100 views per upload → ≈ 20,000 reads/s → cache and CDN are essential
OperationRough latency
L1 cache / main memory reference~1 ns / ~100 ns
Read 1 MB sequentially from memory~10 µs
SSD random read~100 µs
Round trip within a data centre~0.5 ms
Read 1 MB sequentially from SSD~1 ms
Disk seek (spinning disk)~10 ms
Round trip between continents~150 ms

Common mistakes

  • Jumping to a solution ("we'll use Kafka and Cassandra") before agreeing on requirements.
  • Designing for Google scale when the numbers say one server is enough. Over-engineering is a red flag too.
  • Naming technologies without saying why. "A cache, because reads are 100× writes and staleness of a few seconds is fine" beats "Redis".
  • Ignoring failures: what happens when this box dies?
  • Monologuing. Check in with the interviewer; they often hint at where they want depth.

See it animated

Big-O complexity

Interview questions

Try answering out loud first, then open the model answer.

Q1What questions do you ask first in a system design interview?beginner

Who the users are and what they need to do (core features), the expected scale (daily users, requests per second, data size), the read/write ratio, latency and availability targets, whether strong consistency matters, and what's explicitly out of scope. These answers decide every later choice.

Q2How do you estimate requests per second from daily users?beginner

Multiply daily users by actions per user per day, divide by about 10⁵ seconds, then multiply by 2–3 for peak. 50 million users × 10 actions ≈ 500 million per day ≈ 5,000 per second on average, roughly 10–15k at peak.

Q3What's the difference between functional and non-functional requirements?beginner

Functional requirements are features: "users can post a message", "users see a feed". Non-functional requirements are qualities: latency under 200 ms, 99.99% availability, durability, consistency, cost. Non-functional requirements usually drive the architecture.

Q4How do you decide what to deep-dive into?intermediate

Go where the risk is: the component that hits the scale limit first, the part with the hardest consistency requirement, or the feature that makes this system different (the feed fan-out in a social network, ID generation in a URL shortener). Ask the interviewer which area interests them.

Quick check

Question 1

1.About how many seconds are in a day, for estimation?

Question 2

2.What should you do before drawing the architecture?

Question 3

3.Which is a non-functional requirement?

0 of 3 answered

Scalability basics
Log inStart free