Interview questions
Distributed systems trade-offs and architecture choices.
The CAP theorem says that when a network partition occurs, a distributed data store must choose between consistency — every read sees the latest write — and availability — every request gets a non-error response. Partition tolerance is not optional in real networks, so the practical choice is CP or AP behaviour during a partition.
A monolith deploys the application as one unit: simple to build, test and run, and usually the right start. Microservices split it into independently deployable services owned by separate teams, which helps scaling organisations and uneven workloads but adds network calls, distributed data, observability and operational complexity. Choose based on team structure and scaling pain, not trends.
A load balancer spreads incoming requests across multiple servers to increase capacity and availability. It routes with algorithms such as round robin, least connections or hashing, removes unhealthy servers using health checks, and can work at layer 4 (TCP) or layer 7 (HTTP), where it can also route by path, header or cookie.
Consistent hashing places both servers and keys on a hash ring; each key belongs to the next server clockwise. When a server is added or removed, only the keys between it and its neighbour move, instead of almost every key as with hash-mod-N. Virtual nodes give each server several positions on the ring to balance load.
Clarify requirements and scale first, then sketch a simple end-to-end design, then deepen the parts that matter. A reliable flow is: functional and non-functional requirements, capacity estimates, API design, data model, high-level architecture, deep dives on bottlenecks, and finally trade-offs, failure modes and how you would monitor the system.
Start free, install the Windows app and run a practice session today.
