Top 50 System Design Interview Questions and How to Answer Them

What happens to your design when a retry arrives after the payment succeeded, or when a cache serves yesterday's permissions? Those questions turn a diagram into an engineering discussion. This guide offers 50 practice prompts for explaining requirements, failure modes, and architectural trade-offs out loud.

Use the employer's preparation materials to establish the actual format. For example, Amazon's SDE II guide includes system design and encourages candidates to ask questions that validate their design. It does not establish a universal format for every employer or level.

The difficulty groups below are our editorial practice sequence, not measured question frequencies or a company's interview bank. The AI prompts are useful when your target role involves retrieval, model serving, or tool execution; ask your recruiter whether those topics belong in your loop.

The System Design Interview Framework

Try a repeatable framework. We call this suggested practice structure RACED; the time allocations are planning aids, not employer rules. Adjust them to the round length and interviewer guidance.

Step	Duration	What to Do
R - Requirements	3-5 min	Clarify functional and non-functional requirements. Ask about scale, latency, consistency, and availability.
A - API Design	3-5 min	Define the key API endpoints. This shows you think about the system from the user's perspective.
C - Capacity Estimation	3-5 min	Back-of-envelope math: DAU, QPS, storage, bandwidth. Shows you think about scale.
E - Entity Design	5-8 min	High-level architecture diagram. Data model, key services, data flow.
D - Deep Dive	20-30 min	Go deep on 2-3 components. This is where the real discussion happens.

The framework is not rigid. Good interviewers will steer the conversation toward the areas they care about most. But having a structure prevents you from spending 20 minutes on requirements clarification and running out of time for the actual design.

Beginner Questions (1-15): Core Fundamentals

Start here to practice caching, load balancing, database design, and API design. "Beginner" describes the starting scope: any of these prompts can become a deep discussion when failure recovery or operational constraints are added.

1. Design a URL Shortener (TinyURL)

Key concepts: Hashing, base62 encoding, read-heavy system, caching, database sharding. Focus on: How do you generate unique short codes? How do you handle collisions? What is the read:write ratio and how does that inform your caching strategy?

2. Design a Rate Limiter

Key concepts: Token bucket, sliding window, distributed rate limiting. Focus on: Where does the limiter sit in your architecture? How do you handle distributed rate limiting across multiple servers?

3. Design a Key-Value Store

Key concepts: Consistent hashing, replication, CAP theorem, conflict resolution. Focus on: What consistency model do you choose and why? How do you handle node failures?

4. Design a Paste Tool (Pastebin)

Key concepts: Object storage, CDN, TTL-based expiration. Focus on: How do you handle large pastes vs. small pastes? What is your storage strategy?

5. Design an API Gateway

Key concepts: Reverse proxy, authentication, rate limiting, request routing, circuit breaking. Focus on: How do you handle service discovery? What happens when a downstream service is slow?

6. Design a Notification System

Key concepts: Message queues, push vs. pull, device tokens, delivery guarantees. Focus on: How do you handle multiple notification channels (push, email, SMS)? How do you prevent notification storms?

7. Design a Task Scheduler (Cron)

Key concepts: Distributed scheduling, idempotent effects, leases, failure recovery. Focus on: What happens when a worker completes an external effect but crashes before recording success? Distinguish task deduplication from guaranteeing its external effects happen once.

8. Design a Logging System

Key concepts: Write-optimized storage, log aggregation, search indexing (ELK stack). Focus on: How do you handle high write throughput? How do you make logs searchable?

9. Design a Content Delivery Network (CDN)

Key concepts: Edge caching, cache invalidation, origin shielding, geographic routing. Focus on: How do you decide what to cache at the edge? How do you handle cache invalidation?

10. Design a Unique ID Generator

Key concepts: Snowflake IDs, UUID trade-offs, clock synchronization, sortable IDs. Focus on: What properties does your ID need (sortable, globally unique, compact)?

11. Design a URL Crawler

Key concepts: BFS/DFS crawling, politeness constraints, URL frontier, deduplication. Focus on: How do you respect robots.txt? How do you prioritize URLs?

12. Design an Autocomplete System

Key concepts: Trie data structure, prefix matching, ranking by frequency, low-latency reads. Focus on: How do you update suggestions in real-time? How do you handle personalization?

13. Design a Leaderboard

Key concepts: Sorted sets (Redis), real-time ranking, pagination. Focus on: How do you handle ties? How do you scale for millions of players?

14. Design a File Upload Service

Key concepts: Multipart upload, resumable uploads, virus scanning, presigned URLs. Focus on: How do you handle large files? S3's multipart documentation illustrates retrying a failed part without restarting the whole upload.

15. Design a Comment System

Key concepts: Nested comments (tree structure), pagination, real-time updates. Focus on: How do you store and retrieve threaded comments efficiently?

Intermediate Questions (16-30): Distributed Systems

Use these exercises to connect services and reason through distributed failure. Match the scope to the role: expand to multiple regions only after the basic delivery contract is clear.

16. Design Instagram

Key concepts: Photo storage, news feed generation, fan-out-on-write vs. fan-out-on-read, CDN for media. Focus on: How do you generate the home feed? What is your fan-out strategy for celebrities vs. regular users?

17. Design Twitter / X

Key concepts: Timeline generation, tweet storage, follow graph, real-time delivery. Focus on: How do you handle users with millions of followers? What is the write path vs. read path?

18. Design a Chat Application (WhatsApp)

Key concepts: WebSocket connections, message ordering, delivery guarantees, end-to-end encryption. Focus on: How do you handle offline users? How do you ensure message ordering in group chats?

19. Design YouTube / Video Streaming

Key concepts: Video transcoding, adaptive bitrate streaming, CDN, recommendation engine. Focus on: How do you handle video upload and processing pipeline? How do you serve video at global scale?

20. Design Google Drive / Dropbox

Key concepts: File sync, conflict resolution, chunked uploads, version history, deduplication. Focus on: How do you detect and sync file changes? How do you handle conflicts?

21. Design a Search Engine

Key concepts: Inverted index, ranking (TF-IDF, PageRank), crawling, query processing. Focus on: How do you build and update the index? How do you rank results?

22. Design a Ride-Sharing System (Uber/Lyft)

Key concepts: Geospatial indexing, real-time matching, ETA calculation, surge pricing. Focus on: How do you match riders with nearby drivers efficiently?

23. Design a Food Delivery Service (DoorDash)

Key concepts: Three-sided marketplace (customer, restaurant, driver), order tracking, dispatch optimization. Focus on: How do you optimize delivery routes? How do you estimate delivery times?

24. Design a Ticket Booking System (Ticketmaster)

Key concepts: Inventory management, distributed locking, high-concurrency writes, queue management. Focus on: How do you prevent overselling? How do you handle flash sales?

25. Design an E-commerce Platform (Amazon)

Key concepts: Product catalog, shopping cart, order processing, payment system, inventory management. Focus on: How do you handle the checkout flow? How do you ensure payment consistency?

26. Design a Social Network Graph

Key concepts: Graph database, friend-of-friend queries, social graph traversal, privacy controls. Focus on: How do you efficiently find mutual connections? How do you generate friend suggestions?

27. Design a Music Streaming Service (Spotify)

Key concepts: Audio streaming, playlist management, recommendation engine, offline sync. Focus on: How do you handle seamless playback? How do you generate personalized playlists?

28. Design a Metrics Collection System

Key concepts: Time-series database, data aggregation, dashboarding, alerting. Focus on: How do you handle high-cardinality metrics? What is your retention and downsampling strategy?

29. Design a News Feed Aggregator

Key concepts: RSS/Atom parsing, content deduplication, ranking, personalization. Focus on: How do you rank content from different sources? How do you handle real-time updates?

30. Design a Payment System (Stripe)

Key concepts: Idempotency, ledger accounting, reconciliation, access controls, webhook delivery. Focus on: How do you prevent duplicate charges during retries? Stripe documents idempotency keys for supported requests; that contract does not make every downstream workflow exactly once.

Advanced Questions (31-42): Scale and Complexity

These exercises increase the scope of distributed-systems and domain-specific reasoning. Use them for a deeper practice session when the target role calls for that expertise; the category is not a company leveling rubric.

31. Design Google Maps

Key concepts: Tile-based map rendering, routing algorithms (Dijkstra/A*), real-time traffic, geospatial indexing. Focus on: How do you compute routes with real-time traffic data?

32. Design a Distributed Cache (Memcached/Redis)

Key concepts: Consistent hashing, cache eviction policies, replication, cache-aside vs. write-through. Focus on: How do you handle cache invalidation at scale?

33. Design a Distributed Message Queue (Kafka)

Key concepts: Partitioning, consumer groups, ordering, transactions, log compaction. Focus on: What ordering contract do consumers observe? Kafka's design documentation distinguishes delivery semantics and the limits of exactly-once processing when writing to external systems.

34. Design a Distributed Database (Spanner/CockroachDB)

Key concepts: Distributed transactions, TrueTime/hybrid logical clocks, consensus (Paxos/Raft), sharding. Focus on: How do you achieve global consistency with acceptable latency?

35. Design a Real-Time Gaming Platform

Key concepts: Low-latency networking, state synchronization, matchmaking, anti-cheat. Focus on: How do you handle network latency for real-time multiplayer?

36. Design a Stock Trading System

Key concepts: Order matching engine, FIFO queue, low-latency processing, audit trail, market data feed. Focus on: How do you match buy/sell orders at microsecond latency?

37. Design a Live Streaming Platform (Twitch)

Key concepts: Ingest servers, transcoding, CDN distribution, real-time chat, low-latency delivery. Focus on: How do you deliver live video with sub-5-second latency?

38. Design a Collaborative Document Editor (Google Docs)

Key concepts: Operational Transformation (OT) or CRDTs, real-time sync, conflict resolution, cursor tracking. Focus on: How do you handle concurrent edits without conflicts?

39. Design a Global DNS System

Key concepts: Hierarchical resolution, caching, anycast routing, DNSSEC. Focus on: How do you ensure low-latency resolution globally?

40. Design a Fraud Detection System

Key concepts: Real-time stream processing, feature engineering, ML model serving, rule engine. Focus on: How do you balance false positives with fraud detection rate?

41. Design a Distributed File System (HDFS/GFS)

Key concepts: Block storage, replication, metadata, fault tolerance. Focus on: How do you recover metadata and handle failed workers? The GFS paper is a primary storage-design example; the MapReduce paper adds a related distributed-computation example.

42. Design a Content Moderation System

Key concepts: ML-based classification, human review queue, appeal workflow, multi-modal detection. Focus on: How do you handle the trade-off between speed and accuracy?

GenAI System Design (43-50): Role-Specific Practice

These exercises add model serving, retrieval, and evaluation to the discussion. They are a practice track for relevant roles, not evidence that every technology company uses a distinct GenAI interview round.

43. Design a ChatGPT-Style Conversational AI Service

Key concepts: LLM serving, KV cache management, streaming responses, conversation history, rate limiting. Focus on: How do you serve an LLM at low latency to millions of concurrent users? How do you manage GPU resources?

44. Design a RAG-Based Q&A System

Key concepts: Document ingestion, chunking strategies, embedding generation, vector database, retrieval pipeline, prompt construction. Focus on: How do you ensure retrieved context is relevant? How do you handle hallucination?

45. Design an AI Coding Assistant (Copilot)

Key concepts: Code context window, inline suggestions, model routing, latency budgets. Focus on: If the exercise assumes a 200 ms suggestion budget, which work can run before a keystroke? How do you handle multi-file context? This budget is hypothetical, not a measured product guarantee.

46. Design an AI Agent System with Planning and Tool Use

Key concepts: Agent loop, tool calling, planning/reasoning, safety guardrails, execution sandboxing. Focus on: How do you handle multi-step plans that fail at step 3? How do you ensure safety?

47. Design an LLM Fine-Tuning Platform

Key concepts: Training data pipeline, GPU cluster management, LoRA/QLoRA, evaluation, model registry. Focus on: How do you manage training jobs across a GPU cluster? How do you evaluate fine-tuned models?

48. Design a Real-Time AI Translation Service

Key concepts: Streaming ASR, machine translation, TTS, latency optimization, language detection. Focus on: How do you achieve real-time translation with acceptable latency?

49. Design an AI-Powered Search Engine

Key concepts: Hybrid search (keyword + semantic), embedding index, re-ranking with LLM, query understanding. Focus on: How do you combine traditional search with semantic understanding?

50. Design a Multi-Modal AI Content Generation Platform

Key concepts: Image generation pipeline, text-to-image models, content safety, model routing, asset storage. Focus on: How do you serve multiple model types (text, image, video) from a single platform?

How to Structure Any System Design Answer

Regardless of the specific question, these principles will improve your performance:

Start with clarifying questions

Never jump straight into drawing boxes. Spend 3-5 minutes asking questions like: What is the expected scale (DAU, QPS)? What are the most important features? Is this read-heavy or write-heavy? What are the latency requirements? What consistency guarantees do we need?

Think out loud

System design interviews evaluate your thought process as much as your solution. Say "I am considering X vs. Y because..." rather than silently drawing a diagram. The interviewer wants to understand your reasoning, not just see your final answer.

Make trade-offs explicit

When you choose a storage model, say why. When you choose eventual consistency, explain which operation can tolerate stale data. For a concrete strong-consistency design, Google's Spanner paper describes globally distributed transactions and time uncertainty.

State your assumptions and calculate

Memorized latency tables are not measurements of your proposed system. Hardware, payload, topology, and load affect the result. Use explicit hypothetical inputs, carry the units, and say what you would measure.

Hypothetical Input	Calculation or Follow-Up
1 million daily users, 20 reads each	20 million reads/day ÷ 86,400 seconds ≈ 231 average reads/second
Assumed peak is 10× average	≈ 2,310 reads/second; validate against observed traffic
1 million new records/day, 1 kB each	≈ 1 GB/day before indexes, replicas, and metadata, using decimal units
End-to-end response budget	Allocate network, storage, computation, and queueing time; measure tail latency

Common Mistakes That Cost Candidates the Round

1. Jumping into the solution too quickly. Clarify requirements before choosing components so your design addresses the actual problem.
2. Designing for Google scale when the question does not require it. If the interviewer says "design for 10,000 users," do not design for 10 billion. Scale your solution to the stated requirements.
3. Using buzzwords without understanding. Saying "we will use Kafka" without explaining why or what problem it solves undermines your credibility. Only reference technologies you can discuss in depth.
4. Ignoring non-functional requirements. Latency, availability, consistency, and security are often more important than features. Address them explicitly.
5. Not going deep enough. Spending the entire 45 minutes at the high-level diagram level is a weak signal. Pick 2-3 components and go deep on the implementation details, data models, and edge cases.

For behavioral interview preparation that complements your system design prep, see our guide to the 40 most common behavioral questions with STAR method answers. For a complete preparation framework, read our interview preparation guide. And for company-specific preparation at FAANG companies, see our FAANG interview prep guide.