Answer framework
Clarify functional needs, SLOs, RTO/RPO, traffic, data model, compliance, budget, and team capability. Then draw the synchronous path and asynchronous branches and explain failure behavior.
Frequent themes
- High availability: Multi-AZ, health checks, replacement, stateless nodes, and redundant entry points.
- Scalability: vertical vs horizontal; scaling is not elasticity unless capacity can also shrink automatically.
- Decoupling: queues absorb backpressure, event buses route events, workflows orchestrate steps.
- Event-driven design: at-least-once delivery requires idempotency, retries, DLQs, and observability.
- Serverless: reduces server operations but not quota, cold-start, retry, state, and cost design.
- DR: choose backup/restore, pilot light, warm standby, or active/active by RTO/RPO and cost.
- Three-tier: separate entry, application, and data layers, then design each for scale, security, and failure.
Interview sequence
- State constraints and the dominant quality attributes.
- Present the main design and data flow.
- Explain scaling, health, state location, and failure handling.
- Cover security, logs, metrics, traces, backup, and DR.
- Make cost, complexity, and alternative trade-offs explicit.
Remember
A strong architecture answer explains not only the happy path, but what fails, how it is observed and recovered, and why that resilience is worth its cost.