Courses / System Design Interview

Course · Beta · Free

System Design Interview

Learn how large systems are built: how they scale from one server to millions of users, store and copy data safely, stay correct when machines and networks fail, and stay fast under load. Then design sixteen real products end to end, from a link shortener to payments, and defend every trade-off.

Beginner to intermediate39 chapters~22 h reading + practice82 animated scenesUpdated 16 Oct 2026
Start chapter 1What System Design Is: From One Server to Millions33 min · 3 animated scenes

0 of 39 chapters done · saved on this device

Try a scene from chapter 1

Animated sceneA request's journey0:25

A cache miss: the server reads the database, replies, and keeps a copy in the cache

Read it as steps

One sale-night request through the load balancer, servers, cache and database.

What you will be able to do

  • Scale a web service from one server to millions of users with load balancers, stateless servers, caches and CDNs.
  • Pick the right database for an access pattern, then replicate and shard it without losing data or creating hot spots.
  • Keep data correct under concurrency and failure with transactions, idempotent retries, consensus and fencing.
  • Build reliable asynchronous systems with queues, pub-sub, sagas and schedulers, handling retries, ordering and back-pressure.
  • Design for failure across servers, zones and regions with replication, failover, rate limits, circuit breakers and monitoring.
  • Design sixteen real products end to end, from a link shortener to payments, and defend every trade-off in a 45-minute round.

Who it is for

Final-year students, freshers and engineers with up to five years of experience preparing for system design rounds at product companies, and anyone curious about how the apps they use every day stay fast and stay up.

You will need

You can write a small program in any language, know what an HTTP request and a status code are, and can read a simple SQL query. No cloud account, no setup, nothing to install.

Pick a path

Interview this week12 chapters · about 6 hoursEstimates, load balancers, databases, caches, queues, rate limiters, then a link shortener, a news feed, group chat and the practice chapter. Do every drill out loud.
Recommended6 weeks · 45 minutes a dayEvery chapter in order, one product per weekday, and a timed mock round every weekend.
Deep10 weeksEverything, every drill twice, and two mock rounds a week with a friend playing the interviewer.

Curriculum

4 modules · 39 chapters · 22 h
Module 1Foundations4 chapters · 2 h 9 min0 of 4 done

How a system grows from one server to millions, how to size it with quick estimates, and the properties every design is judged on: availability, consistency, latency and cost.

  1. Not started: 01What System Design Is: From One Server to Millions33 min · 3 scenesHow a product grows from one server to millions, one bottleneck at a time, and how a first design starts
  2. Not started: 02Back-of-the-Envelope Estimates32 min · 3 scenesTurn users into peak load, servers, storage and a cost line, sized for a food app on a cricket-final night
  3. Not started: 03Abstractions That Hide the Network32 min · 1 sceneWhy remote calls fail or run twice, how to retry them safely, and which consistency each operation needs
  4. Not started: 04Availability, Reliability, Scalability and Latency32 min · 2 scenesHow to turn availability, durability, scale and speed into numbers, and the designs that meet them
Module 2Building Blocks17 chapters · 8 h 41 min0 of 17 done

The seventeen parts large systems are built from: load balancers, databases, caches, queues, consensus and more. How each works inside, what it costs, and how it fails.

  1. Not started: 05DNS: How a Name Becomes an Address29 min · 2 scenesHow names become addresses through resolvers and TTL caches, and how to move, fail over and defend a domain
  2. Not started: 06Load Balancers32 min · 2 scenesHow load balancers spread traffic, drop dead servers in seconds, and deploy without losing a request
  3. Not started: 07Databases: Replication, Partitioning and Choosing a Store33 min · 3 scenesHow to pick a store, index it, keep it correct, and replicate and partition it without losing data
  4. Not started: 08Consensus, Leader Election and Coordination32 min · 2 scenesHow a majority vote, numbered terms, leases and fencing tokens keep exactly one machine in charge
  5. Not started: 09Designing a Key-Value Store32 min · 2 scenesHow a store spreads billions of keys on a ring, keeps three copies and stays correct as machines fail
  6. Not started: 10Content Delivery Networks31 min · 2 scenesHow edge caches cut latency and origin load, and how cache keys, TTLs, purges and shields keep them correct
  7. Not started: 11Unique IDs and Sequencers27 min · 1 sceneHow to make IDs that never collide and sort by time, from UUIDs to 64-bit IDs and gap-free invoices
  8. Not started: 12Monitoring: Metrics, Logs, Traces and Alerts32 min · 2 scenesHow golden signals, symptom alerts, traces and error budgets show when users hurt, and how to scale them
  9. Not started: 13Distributed Caches33 min · 3 scenesHow a shared cache absorbs reads, stays fresh when replicas lag, and survives stampedes, hot keys and restarts
  10. Not started: 14Message Queues34 min · 3 scenesHow a queue absorbs spikes, why delivery is at least once, and how a replicated log broker keeps messages
  11. Not started: 15Publish-Subscribe27 min · 2 scenesHow one event reaches every service that cares, how subscribers survive repeats, and how sagas span services
  12. Not started: 16Rate Limiters33 min · 3 scenesHow to cap request rates fairly across many gateways, with no races, and what to do when the limiter fails
  13. Not started: 17Blob Stores31 min · 1 sceneHow files are chunked, copied or erasure-coded across racks, repaired fast, and served with signed URLs
  14. Not started: 18Distributed Search31 min · 1 sceneHow an inverted index finds and ranks matches, and how shards, replicas and hedging keep queries fast
  15. Not started: 19Distributed Logging26 min · 1 sceneHow agents, a queue and tiered storage turn lines from hundreds of machines into one cheap, safe search
  16. Not started: 20Distributed Task Schedulers32 min · 2 scenesHow leases, idempotent tasks and sharded minute buckets run a billion timed tasks without losing one
  17. Not started: 21Sharded Counters26 min · 1 sceneHow splitting one hot counter into many rows absorbs a viral burst, and when a sketch counts better
Module 3Design Real Products16 chapters · 10 h 34 min0 of 16 done

Sixteen real products designed end to end: requirements, estimates, the data model, the read and write paths, the hard parts in depth, and what breaks at scale.

  1. Not started: 22Design a Link Shortener39 min · 2 scenes · interview drillHow to mint unique 7-character codes, serve 23,000 redirects a second from cache, and why 302 beats 301
  2. Not started: 23Design a Video Streaming Platform40 min · 3 scenes · interview drillHow one upload becomes segments any phone can play, and how CDNs and adaptive players carry 20 Tbps
  3. Not started: 24Design a Q&A Site39 min · 2 scenes · interview drillHow to serve 100 million page views a day from cached pages, count every vote once and keep a viral page cheap
  4. Not started: 25Design Maps and Navigation40 min · 2 scenes · interview drillHow a maps app serves tiles, routes 150 km in milliseconds with shortcuts, and turns phone pings into ETAs
  5. Not started: 26Design Nearby Places Search35 min · 2 scenes · interview drillHow geohash cells, nine-cell searches and per-cell caches find every café near you, even across a border
  6. Not started: 27Design a Ride-Sharing App42 min · 2 scenes · interview drillHow a million moving drivers are tracked in memory, matched with expiring offers, and never double booked
  7. Not started: 28Design a Microblog39 min · 2 scenes · interview drillHow fan-out on write, read-time merging for huge accounts, and a sized timeline cache keep timelines fast
  8. Not started: 29Design a News Feed38 min · 2 scenes · interview drillHow a feed gathers candidates, scores them once per session, and pages with cursors that never repeat posts
  9. Not started: 30Design a Photo-Sharing App37 min · 2 scenes · interview drillHow photos reach the blob store without touching your servers, pass a safety scan, and stay cheap to store
  10. Not started: 31Design a Group Chat App42 min · 3 scenes · interview drillHow a chat app keeps one order per conversation and never loses a sent message, online or off
  11. Not started: 32Design a Web Crawler41 min · 2 scenes · interview drillHow a URL frontier fetches 100 M pages a day, finds what matters first and never overloads a site
  12. Not started: 33Design Typeahead Suggestions40 min · 4 scenes · interview drillHow a trie with stored top-10 lists, edge caches and an hourly build answer each keystroke within 100 ms
  13. Not started: 34Design Collaborative Document Editing40 min · 2 scenes · interview drillHow one owner per document, operational transformation and CRDTs keep every copy of a shared doc identical
  14. Not started: 35Design a Notification System39 min · 3 scenes · interview drillHow one service sends OTPs, order updates and offers on time, once, even when providers fail
  15. Not started: 36Design a Ticket Booking System41 min · 2 scenes · interview drillSell 40,000 seats to 10 lakh fans once each: conditional updates, expiring holds and a fair waiting room
  16. Not started: 37Design a Payment System42 min · 3 scenes · interview drillTake a ₹1,000 UPI payment without charging twice: idempotency keys, a state machine, a ledger, reconciliation
Module 4Wrap-Up2 chapters · 58 min0 of 2 done

What real outages teach about designing for failure, and how to practise until designing a system on a whiteboard feels routine.

  1. Not started: 38Lessons from Real Outages32 min · 2 scenesHow small changes become big outages, and how cells, canaries, retry limits and tested restores contain them
  2. Not started: 39How to Practise: A 45-Minute Checklist26 minHow the design round is judged and run on a 40-minute clock, one worked round, and a six-week practice plan

How this course is made

Written by the aiiolar team. Every chapter is fact-checked by a second reviewer, checked against a shared set of numbers so chapters agree, and checked for originality before it goes up. Found a mistake? Tell us and we fix it within a week.

Questions

Is it really free?

Yes. Every chapter, scene and drill is free, with no sign-up wall.

Do I need to sign in?

No. Your progress is saved on this device. Sign in only if you want it on your phone and your laptop.

Will it work on slow mobile data?

Yes. Text loads first and scenes load only when you reach them, and every scene also has a text version you can read instead.

Is this enough for my interview?

It prepares you for the high-level design round. Many companies also run a low-level design or machine-coding round; the last chapter explains how they differ and what carries over.

How is this different from reading blog posts?

Every product chapter makes you attempt the design first, then shows a timed 45-minute answer and drills that change the requirements, the way interviewers do.

Who checks the content?

Each chapter is fact-checked by a second reviewer, checked against a shared set of numbers so chapters agree, and checked for originality before it goes up. Found a mistake? Tell us and we fix it.