Architecture Interview Guide

Define software architecture in your own words. What is in it, what is not, and how would you tell whether a system has good architecture?

What is the difference between code architecture and system architecture? Can one be good while the other is bad? Give an example of each direction.

A candidate draws a load balancer, a cache, a queue and a CDN before writing down a single number. How should each of those be justified? Walk me through a system that grows from browser → backend → database, adding components only when a measured problem appears.

A team of eight engineers runs a monolith with 2,000 requests per second and wants to move to microservices. What would you ask them, and when is the move justified?

Defend or attack the statement "a monolith is bad architecture". How does a monolith scale, and where is its real ceiling?

A modular monolith is only modular if the boundaries hold. What mechanisms actually keep the Order module from reaching into the Payment module's tables, and how does that prepare a later extraction?

You are splitting a system into services. What makes a boundary good, what makes it bad, and how would you find the right one for orders, inventory and payments?

Why do microservices insist on a database per service? What do you lose, and how do you answer a question that needs data from two services?

Compare layered architecture with hexagonal (ports and adapters). What problem does hexagonal solve that layers do not, and when is the difference not worth the extra structure?

Clean Architecture says source code dependencies must point inward, toward the domain. Why that direction? What goes wrong when it is broken, and how do you enforce it?

How do you recognise that a codebase has more abstraction than it needs, and how do you decide what to remove?

What is the difference between an event and a command, and why does it matter which one you send between services?

A product manager asks for "exactly-once" processing of payment events. Is that possible? What do you actually build?

Consumers sometimes see `OrderUpdated` before `OrderCreated`. Why does that happen in a partitioned log, and how do you choose a partition key?

When would you choose a work queue, a pub/sub topic, or a Kafka-style log? What does each guarantee about who receives a message and for how long?

How do you decide whether an interaction between two services should be a synchronous request or an asynchronous message? Give the failure modes of both.

When would you choose GraphQL over REST for an API, and what does GraphQL make harder?

Teams often use gRPC between internal services and REST at the edge. What does gRPC buy internally, what does it cost, and why not expose it to browsers?

A dashboard needs live updates. Compare polling, long polling, server-sent events and WebSockets. What does each cost at 100,000 connected clients?

What should an API gateway do, what must it never do, and how do you stop it from becoming the new monolith?

You need to change a public API in a way that breaks existing clients. How do you version it, how long do you support old versions, and how do you avoid needing a new version at all?

Why does horizontal scaling require stateless services? What state is unavoidable, and how do you handle it?

Compare round robin, least connections and consistent hashing. When does each fail, and what do health checks get wrong?

Design the invalidation strategy for a product catalogue cache in front of Postgres. Cover TTL, explicit invalidation, stampedes and what "stale" is allowed to mean.

What should you put on a CDN, what should you not, and how do you invalidate it? Include the numbers that justify a CDN in the first place.

A single-server app with one Postgres database is slowing down as traffic grows. Walk me through what you do first, second and third, and what evidence moves you from one step to the next.

A mobile client calls `POST /payments`, the gateway times out after 30 s, and the client retries. How do you guarantee the customer is charged once?

An email-sending worker fails on a transient SMTP error. Describe a retry policy, and what happens to a job that never succeeds.

A pipeline ingests 100,000 events per second but the consumers process 20,000. The queue grows. What are your options and how do you choose?

Placing an order must create the order, charge the card and reserve stock across three services with their own databases. How do you make that consistent?

In an order saga, the payment step fails after inventory has been reserved. What runs, what state does the user see, and what can go wrong during compensation itself?

A team wants to introduce CQRS for their SaaS product. What problem must exist for that to be a good decision, and what will it cost them?

A bank-ledger team stores `AccountOpened +100`, `PaymentMade −20`, `RefundReceived +20` and replays them for the balance. What does this buy them, what does it cost, and where would you not do it?

A payment provider slows down. Ten minutes later it is completely down and so is your checkout. Your services all retry on failure. Explain what happened and how to retry safely.

Walk me through Closed, Open and Half-Open. What triggers each transition, and what should the caller do while the breaker is open?

You need to limit an API to 100 requests per minute per client. Compare fixed window, sliding window and token bucket, pick one, and explain how it works across 20 gateway instances.

You own a checkout API. Define SLIs and an SLO, compute the error budget, and explain how the budget changes what the team does.

How much downtime per year is 99.9%? If your service calls three dependencies that are each 99.9%, what availability can you promise?

You shard a cache across 10 nodes with `hash(key) mod 10`. You add an eleventh node. What happens, and how does consistent hashing fix it?

Explain CAP without the "pick two" slogan. For a multi-region user-profile store, what happens during a partition under each choice, and how do quorums fit in?

Instances autoscale, get rescheduled and roll out several times a day. How does Service A find a healthy instance of Service B, and what goes wrong with stale information?

Explain the difference between logs, metrics and traces, and for a "checkout is slow" incident, which one you would open first and why.

A trace shows: Gateway 20 ms → User Service 35 ms → Order Service 180 ms, and inside Order Service a Database span of 140 ms. What do you conclude, what do you check next, and how did the trace id get from the gateway to the database span?

Design a service like bit.ly. Walk me through requirements, scale, the API, the data model, and how the design evolves as traffic grows.

Design a service that sends email, SMS and push notifications for the rest of the company. Cover the API, the pipeline, retries, rate limits, deduplication and how you know a notification was delivered.

Design real-time one-to-one and group chat with online/offline presence. Cover connections, message routing, ordering, delivery receipts, presence at scale, and offline users.