Start Here¶
What system design actually is¶
Writing code solves a problem for one request. System design solves it for millions of requests, all at once, while machines fail, networks slow down, and data keeps growing.
So a system design question is really four questions stacked on top of each other:
- What must the system do? (functional requirements — "users can shorten a URL")
- How well must it do it? (non-functional — latency, availability, consistency, cost)
- How big is it? (scale — users, requests per second, bytes stored)
- What breaks first, and what do we do about it? (bottlenecks, failures, trade-offs)
Every chapter on this site answers those four questions for one system or one component.
flowchart LR
A[Vague problem<br/>'Design YouTube'] --> B[Clarify scope<br/>features + scale]
B --> C[Estimate<br/>QPS, storage, bandwidth]
C --> D[High-level design<br/>boxes + arrows]
D --> E[Deep dive<br/>the 2-3 hard parts]
E --> F[Trade-offs,<br/>failures, wrap-up]
How to use this site¶
The 3-pass method
- Pass 1 — understand. Read the chapter top to bottom. Draw every diagram yourself on paper before looking at it closely. Play with the interactive widgets where they exist.
- Pass 2 — reproduce. Close the page. Set a 35-minute timer and design the system from scratch on a whiteboard or paper, talking out loud. Then compare with the chapter.
- Pass 3 — defend. Open the Interviewer follow-ups section and answer each question out loud before expanding the answer.
English or Hinglish? Both versions carry the same depth. Use whichever helps you think. A useful trick: read in Hinglish to build intuition, then re-read the English version so you can say it fluently in an interview.
Suggested 4-week plan¶
| Week | Chapters | Goal |
|---|---|---|
| 1 | 01 Scale · 02 Estimation · 03 Framework | Vocabulary + the method. You should be able to draw the "millions of users" architecture from memory and estimate QPS/storage in under 3 minutes. |
| 2 | 04 Rate Limiter · 05 Consistent Hashing · 06 KV Store · 07 Unique IDs | The building blocks. These show up inside every later design — don't rush them. |
| 3 | 08 URL Shortener · 09 Crawler · 10 Notifications · 11 News Feed | First four case studies. Do each one timed (pass 2) at least once. |
| 4 | 12 Chat · 13 Autocomplete · 14 YouTube · 15 Google Drive · 16 Cheat Sheet | Last four case studies + revision. End with two full mock interviews. |
If you have only one week: do 01, 02, 03 fully, skim 05 and 06, then do 08, 11, 12 and 14 timed.
The vocabulary you need on day one¶
You'll meet these words constantly. Each gets a proper explanation later, but here's the one-liner. The Glossary has 100+ more.
| Term | One-liner |
|---|---|
| Latency | How long one request takes (ms). |
| Throughput / QPS | How many requests the system handles per second. |
| Availability | Fraction of time the system is up and answering (99.9% = ~8.8 h downtime/year). |
| Scalability | Ability to handle more load by adding resources. |
| Horizontal vs vertical scaling | More machines vs a bigger machine. |
| Replication | Keeping copies of the same data on several machines. |
| Sharding / partitioning | Splitting data so each machine holds only a part. |
| Cache | A fast, small, temporary store that holds hot data to avoid slow work. |
| Load balancer | Spreads incoming requests across many servers. |
| Message queue | A buffer that lets producers and consumers work at different speeds. |
| Consistency | Whether every reader sees the latest write. |
| Idempotency | Doing the same operation twice has the same effect as doing it once. |
A note on Java code¶
Code samples are in Java (the most common interview language). They're written to be readable and correct, not production-hardened — think of them as executable explanations.
Credits¶
This site is a companion to Alex Xu, System Design Interview – An Insider's Guide (2nd ed.). The chapter order follows the book. All prose, diagrams, code and examples here are original; nothing is copied from the book. Please buy it — it's worth it.