# Backend Patterns > Backend architecture patterns and best practices for scalable server-side applications. Parent: [Software Engineering Patterns](https://llms-explorer.com/tree/software-engineering-patterns/) · 7 facets · 25 facts · page: https://llms-explorer.com/tree/backend-patterns/ ## Backend Development Patterns - Backend architecture patterns and best practices for scalable server-side applications. — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#backend-development-patterns) ## When to Activate - Designing REST or GraphQL API endpoints — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#when-to-activate) - Implementing repository, service, or controller layers — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#when-to-activate) - Optimizing database queries (N+1, indexing, connection pooling) — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#when-to-activate) - Adding caching (Redis, in-memory, HTTP cache headers) — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#when-to-activate) - Setting up background jobs or async processing — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#when-to-activate) - Structuring error handling and validation for APIs — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#when-to-activate) - Building middleware (auth, logging, rate limiting) — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#when-to-activate) ## Structured Logging - Remember: Backend patterns enable scalable, maintainable server-side applications. Choose patterns that fit your complexity level. — [source](https://llms-explorer.com/sources/mdb-context-hub/backend-patterns/#structured-logging) ## Where this helps - Designing a new REST or GraphQL API's endpoint and layering conventions before the first controller, service, or repository file is written. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Diagnosing a database-bound slow endpoint by spotting N+1 query patterns, missing indexes, or an exhausted connection pool. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Deciding where caching belongs in a request path — Redis, in-memory, or HTTP cache headers — to cut load on a hot read endpoint. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Structuring background or async processing so long-running work such as emails, exports, or webhooks doesn't block the request-response cycle. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Project ideas - Build a small service with clean repository/service/controller separation and compare it against a version with logic mixed directly into route handlers to feel the maintainability difference. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Add a request-scoped middleware chain for auth checking, structured request logging, and rate limiting to an existing API and measure the added latency per middleware. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Introduce a Redis cache layer in front of a read-heavy endpoint with explicit invalidation on writes, and load-test before and after. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Move a slow synchronous operation, such as sending an email on signup, into a background job queue and confirm the API response time drops. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Antipatterns - Fetching a collection and then looping over it to fetch each related record individually, the N+1 query problem, instead of a single joined or batched query. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Skipping connection pooling and opening a new database connection per request, which caps throughput under load far below what the database can actually handle. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Putting business logic directly in route or controller handlers instead of a service layer, making the same logic impossible to reuse or unit-test in isolation. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Adding caching without an invalidation strategy, so stale data silently persists past the point where the underlying record changed. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Known issues - Repository/service/controller layering adds indirection that is overkill for a small CRUD service; the pattern earns its cost mainly once a codebase has real complexity to isolate. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - HTTP cache headers only help when clients and intermediate proxies actually respect them — a misconfigured Cache-Control header can serve stale data to every downstream consumer at once. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Background job queues introduce their own failure mode: a stuck or crash-looping worker can silently stop processing while the API layer still reports success. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Rate-limiting middleware implemented per-instance with in-memory counters doesn't work correctly once the service runs behind a load balancer with multiple instances; it needs a shared store. — [source](https://llms-explorer.com/tree/backend-patterns/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Context files - [Backend Patterns](https://llms-explorer.com/downloads/sources/mdb-context-hub/backend-patterns.md)