The map

Not started
In progress
Done
Builds on

01The runtimeMemory, collectors, threads

02A request's journeyContainer, filters, security, transactions

03Data and testsQueries, pools, tests across boundaries

04Running it in productionSignals, safe releases, model calls

05Next: system designSame ideas, at service scale

TimeoutLibrary ↗Every outgoing call holds a thread and often a connection until it returns. A deadline on each one is what gives them back.

Circuit breakerLibrary ↗When a dependency, a model provider included, keeps failing, stop calling it for a while and serve a fallback instead of queueing threads behind it.

Idempotency keysLibrary ↗Clients and queues retry. An idempotency key lets the service accept the same request twice and act once.

Cache-asideLibrary ↗The usual first cache in front of a database: look in the cache, otherwise load and store. @Cacheable is this pattern behind a proxy.

Transactional outboxLibrary ↗Writing to the database and publishing an event in one step without two-phase commit: write both in the same transaction, publish afterwards.

Competing consumersLibrary ↗Background jobs taken from one queue by many workers: how a backend scales the work that doesn't belong in a request.

01The runtimeMemory, collectors, threads0/4
02A request's journeyContainer, filters, security, transactions0/5
03Data and testsQueries, pools, tests across boundaries0/5
04Running it in productionSignals, safe releases, model calls0/4
05Next: system designSame ideas, at service scale

This role uses the JVM and Spring Boot as its stack because they show every layer of a request explicitly. If your services are in Node.js, the Node path covers the same runtime ideas (the event loop instead of thread pools, streams and backpressure, worker threads), and the later stages apply unchanged. Topics are shared with the Java, Spring Boot, QA automation and AI engineering paths, with shared progress.

Backend engineer: from a request to a service you can run

The map

Where this path leads