Technology
19 min read
Design the payment platform behind checkout and refunds, where correctness matters more than speed: payment intents and attempts, an explicit state machine, idempotency at every hop, asynchronous confirmation through verified webhooks, a double-entry ledger with a runnable example, reconciliation against provider reports, refunds, provider failover, fraud and PCI scope, and an audit trail.
Technology
16 min read
Design ticketing for concerts and matches where thousands of people want the same seats at the same moment: seat inventory modelling, holds with expiry, atomic all-or-nothing seat claims, idempotent payment and confirmation, what happens when payment succeeds after a hold expires, virtual waiting rooms for big on-sales, hot inventory, bots, failures and the invariant that a seat is never sold twice.
Technology
17 min read
Design a service that runs millions of one-time and recurring jobs at the right time, at least once, across many machines: time-bucketed storage, dispatchers that own partitions through leases, atomic claims that prevent double dispatch, queues and workers with heartbeats, retries and dead letters, misfire policies after outages, top-of-the-hour spikes, time zones, and safe HTTP callbacks.
Technology
14 min read
Design the suggestions that appear as users type in a search box: requirements and estimates per keystroke, why querying the database per keystroke fails, an offline pipeline that turns search logs into prefix-to-top-K tables or tries, in-memory serving, client debouncing and CDN caching, trending updates, sharding and replication, privacy filtering and failures.
Technology
16 min read
Design a home feed that shows posts from accounts a user follows: estimates for posting, reading and fan-out, fan-out on read versus fan-out on write, the hybrid model that handles celebrity accounts, precomputed feed storage, hydration, ranking with a chronological fallback, cursor pagination, deletions and privacy, and what changes at social-media scale.
Technology
18 min read
Design one-to-one and group chat for tens of millions of concurrent users: WebSocket connection gateways, a session registry for routing messages to the right server, per-conversation sequence numbers for ordering, acknowledgements and deduplication, offline delivery with push and sync, multi-device cursors, presence, partitioned message storage, failures and security.
Technology
17 min read
Design a notification platform that other services use to reach users by push, SMS, email and in-app inbox: requirements and estimates, a 202-accepted API with idempotency keys, preferences and templates, priority queues per channel, provider rate limits and failover, retries and dead-letter queues, deduplication, delivery tracking, campaign fan-out, failures, security and observability.
Technology
15 min read
Design an in-memory cache service shared by many applications: requirements and estimates, the get/set/delete API, a single-node design with TTLs and LRU eviction, then partitioning with consistent hashing or hash slots, client-side versus proxy routing, replication and failover, rebalancing, hot keys, cache stampedes, and what consistency a cache can and cannot promise.
Technology
19 min read
Design a rate limiter that enforces limits across many API servers: requirements, estimates, where the limiter sits, token bucket, leaky bucket, fixed window, sliding log and sliding window counter algorithms with a runnable comparison, atomic Redis implementations, local plus global hybrids, failure behaviour (fail open or fail closed), multi-region limits and response headers.
Technology
20 min read
The classic first system design problem, worked end to end with the course's method: requirements, estimates, APIs, data model, a simple baseline, short-code generation and collisions, 301 versus 302 redirects, caching a read-heavy path, expiry, analytics off the critical path, sharding, failures, abuse prevention, and the Pastebin variant with object storage.
Technology
14 min read
A practical, repeatable way to run a 45-minute system design interview: what interviewers typically evaluate, a 21-item checklist grouped into seven phases, a time plan, phrases for clarifying, estimating and stating trade-offs, how to handle hints and pushback, how SDE-1 and SDE-2 expectations differ, and a compact worked example.
Technology
16 min read
The security decisions that shape an architecture: trust boundaries and least privilege, authentication with passwords, MFA and identity providers, sessions versus tokens and JWTs, object-level authorisation, service-to-service identity, encryption in transit and at rest, secrets management, input validation and abuse prevention, SSRF, data privacy, audit logs and payment data.
Technology
14 min read
When services communicate by publishing facts instead of calling each other: commands versus events, event notification versus event-carried state transfer, designing and versioning events, ordering, replay, and the two patterns often paired with events: CQRS read models and event sourcing, with honest guidance on when each is worth its complexity.
Technology
14 min read
An honest comparison of monoliths, modular monoliths and microservices: what each gives and costs, why a modular monolith is often the best start, how to draw service boundaries and own data, synchronous versus asynchronous communication, API gateways, service discovery, service meshes, the distributed-monolith trap, and migrating with the strangler fig pattern.
Technology
14 min read
If you cannot observe a system, you cannot operate it. Learn the three signals (structured logs, metrics and distributed traces), the golden signals and RED and USE methods, percentiles and histograms, trace context propagation and OpenTelemetry, SLIs, SLOs and error-budget alerts, health checks and synthetic monitoring, and how to observe asynchronous systems without drowning in data.
Technology
14 min read
How systems survive failures of servers, availability zones and whole regions, and recover from data disasters. Covers failure domains, redundancy and failover, availability arithmetic with correlated failures, RPO and RTO, the four disaster-recovery strategies from backup-and-restore to active-active, multi-region write strategies, backups that actually restore, and safe deployments.
Technology
16 min read
How one slow dependency takes down a whole system, and the techniques that stop it: timeouts and deadlines, retry budgets, circuit breakers, bulkheads, rate limiting, load shedding, backpressure and graceful degradation, with a runnable circuit breaker and a degradation plan for a real page.
Technology
16 min read
When only one node may do something at a time: when you actually need a distributed lock and when you do not, locks for efficiency versus correctness, leases and why they need fencing tokens, Redis and database locks and their limits, leader election, and consensus (Raft, majorities, ZooKeeper and etcd) explained at interview depth.
