Concept-level guides that explain when to use each pattern, where it breaks, and how to defend it in an interview.
Backpressure mechanisms slow, shed, or buffer incoming work when consumers cannot keep up.
Bloom filters quickly test possible set membership with false positives but no false negatives.
CDC reads transactional logs (WAL/binlog/oplog) and emits ordered change events for consumers.
Circuit breakers monitor downstream errors/latency and temporarily block requests when dependency health degrades.
Consistent hashing places nodes and keys on a hash ring so membership changes move only a bounded subset of keys.
A DLQ stores events that exceed retry policy so the main stream can continue while failures are investigated.
Event sourcing stores immutable domain events as source of truth; current state is materialized from the event stream.
Practical exactly-once combines at-least-once delivery with idempotent writes and dedup/transactional boundaries.
Fanout-on-write pushes updates to follower timelines proactively; fanout-on-read computes results during request.
Active-active geo replication allows multiple regions to accept reads/writes with conflict resolution and convergence.
An idempotency key uniquely identifies a client operation so duplicate requests can return the original result.
Leader election protocols choose one active node to coordinate tasks and transfer leadership on failure.
OCC allows concurrent updates and rejects stale writes when version checks fail.
Quorum consistency requires operations to be acknowledged by a subset of replicas such that read and write quorums overlap.
Read-through caching serves reads from cache and automatically fetches from the source of truth on cache miss.
Saga pattern breaks distributed transactions into ordered local steps with compensation for failures.
Sharding splits datasets and traffic across multiple partitions by hash, range, tenant, or hybrid keys.
Token bucket refills tokens at fixed rate and permits requests while tokens remain, enabling bursts up to bucket capacity.
The outbox pattern writes both domain mutation and event row in the same transaction to avoid dual-write inconsistency.
Write-behind caching accepts writes into a fast cache and persists to durable storage asynchronously in batches.