Messaging & Event-Driven Architecture

Decouple components by communicating through messages instead of synchronous calls. Producers don’t know (or wait for) consumers — which buys scalability, resilience, and independent evolution, at the cost of eventual consistency and harder debugging.

Queues vs Pub/Sub vs Streams

• Queue (point-to-point) — one message, one consumer; work distribution. (SQS)
• Pub/Sub (fan-out) — one event, many subscribers; notification. (SNS, EventBridge)
• Log/Stream — ordered, replayable, multiple independent consumers with their own offsets. (Kinesis, Kafka)

flowchart LR
  P[Producer] --> T((Topic / Bus))
  T --> A[Service A]
  T --> B[Service B]
  T --> S[Stream consumer]

Delivery Semantics

• At-most-once — may drop, never duplicate.
• At-least-once — never drops, may duplicate → consumers must be idempotent.
• Exactly-once — hard/expensive; usually achieved as at-least-once + dedup.

Core Patterns

• Event notification — “something happened”; thin event, consumer fetches details.
• Event-carried state transfer — fat event carries the data; consumers avoid callbacks.
• Event sourcing — store the sequence of events as the source of truth; rebuild state by replay.
• CQRS — separate write model from read model(s); pairs naturally with event sourcing.
• Saga — coordinate a distributed transaction as a series of local steps + compensating actions (choreography or orchestration via Step Functions).
• Outbox — write the event and the DB change in one local transaction; a relay publishes it — fixes the dual-write problem.

Reliability Concerns

• Idempotency keys — dedupe at-least-once delivery.
• Dead-letter queues — quarantine poison messages for inspection/replay.
• Ordering — only guaranteed within a partition/message-group; design so global order isn’t required.
• Back-pressure — queues smooth spikes but can hide a consumer that’s permanently too slow — alarm on queue depth and age.

Choosing on AWS

SQS (work queue) · SNS (fan-out) · EventBridge (routing/SaaS/schema) · Kinesis (ordered replayable stream) · Step Functions (orchestration).