System design interviews
Interview guides & deep dives
Free, readable guides for classic system design questions—how interviewers phrase them, what strong answers include, tradeoffs, and FAQs. Each guide links to the matching interactive practice problem with AI feedback.
Guides are curated for clarity and interview realism. For fundamentals (CAP, sharding, caching, etc.), see system design topics.
Facebook Feed System Design
Your friend posted ten minutes ago but your feed still looks like yesterday—fan-out lag or ranker timeout is the product failure. Hybrid fan-out, celebrity policy, and degraded ranking are what interviewers grade.
Instagram Feed System Design
Reels and photos share one home feed but media bytes dwarf text metadata—ranking must not block on transcoding, and a influencer post cannot fan out 50M mailbox writes. Media pipeline + hybrid feed is the Instagram-shape
Rate Limiter System Design
An API gateway must enforce per-client limits (by user, IP, API key, or route) to protect backends and keep usage fair. The interesting part is doing it at millions of checks per second with sub-millisecond overhead, cor
Distributed Cache System Design
Flash sale starts and one product id owns a Redis shard—CPU looks fine while the database melts. The interview is consistent hashing, eviction, replication lag, and cache-aside invalidation, not memorizing RDB flags.
Parking Lot System Design
Saturday mall rush: entry gates disagree with payment kiosks on whether level 3 is full because spot counts updated eventually. The OOD parking lot interview still needs concurrency on spots, ticket state, and payment—at
Job Scheduler System Design
Midnight cron fires 50K jobs at once and your cluster oversubscribes CPUs while a payment reconciliation job waits behind a low-priority email batch. Scheduling, priority, retries, and exactly-once side effects define th
WhatsApp System Design
Imagine billions of people treating chat like a utility—one tap to send, one glance at delivery ticks, and trust that no one at the company can read the thread. Within seconds of a viral moment, a family group shares a v
Dropbox System Design
A user uploads a 10 GB video on hotel Wi-Fi—the upload must resume after disconnect while another device still shows "syncing" with the wrong version until metadata catches up. Chunked storage, revision sync, and conflic
Nearby Friends System Design
Friend map updates every 30 seconds for 10M users in a city—naive "query all friends distance" is O(n) per pan. Geospatial index + privacy grid + push diffs keep the feature usable without draining batteries.
Airbnb Reservation System Design
Guest pays for July 4 weekend while host calendar still shows open on another channel—channel manager sync and atomic date-range booking prevent double-booking nightmares. Focus on reservation core, not full marketplace
Vaccine Booking System Design
Imagine millions of people trying to book a vaccine appointment the moment registration opens. Within seconds, the system receives an enormous spike in traffic, available slots disappear instantly, and thousands of users
Booking Waitlist System Design
Concert sells out in 90 seconds but 200K fans stay on waitlist hoping for cancellations—when one seat frees, five notification workers must not sell it twice. Waitlist queue, hold TTL, and fair notify order are the produ
Notification System Design
Flash sale push goes to 40M devices in one minute—APNs/FCM rate limits bite while email backlog grows in a queue that loses ordering for password resets. Multi-channel delivery, preferences, and idempotent templates sepa
Facebook Post Search System Design
User searches their timeline for "beach wedding 2019" across posts with photos, links, and privacy scopes—global search must never return friends-only posts to strangers. Per-user index shard + privacy filter on every qu
Auth System Design
Credential stuffing hits 50K logins/min with reused passwords while legitimate users on mobile need refresh tokens that survive app restarts without keeping JWTs forever. Sessions, password hashing, and MFA hooks are the
Logging System Design
Incident at 2am: 400 TB/day of logs and grep times out—SREs need tail live and search last 7 days without one Elasticsearch cluster dying from a rogue * query. Ingestion, indexing, and retention tiers are the interview.
Practice with structured feedback
Open the full catalog to pick any problem, or jump in from a guide above.
Browse practice problems