engineering-system-design

Start by gathering functional and non-functional requirements such as scale, latency, and availability, then produce a high-level design with component diagrams, data flow, and API contracts. Follow with deep dives into data models, caching, and error handling, and document trade-offs explicitly.

The choice depends on your consumers and data patterns: REST suits simple resource-based APIs, GraphQL fits clients needing flexible queries, and gRPC works best for high-performance internal service communication. Evaluate each against team familiarity, tooling, and latency requirements.

A design document should include requirements, a component diagram, data flow, API contracts, storage choices, scaling and failover strategy, and explicit trade-off analysis. It should also state assumptions and identify what to revisit as the system grows.

Vertical scaling is simpler and suits early-stage systems with modest load, while horizontal scaling supports higher availability and growth at the cost of added complexity in state management and failover. Estimate load first, then choose based on growth projections and redundancy needs.

A structured framework guides reasoning but cannot replace domain-specific benchmarks, prototyping, or load testing. Decisions like storage choice and caching strategy still require validation against real traffic patterns and infrastructure constraints.