Event-driven patterns with EventBridge

Deploying Applications on AWS

Dunieski Otano

Amazon Web Services Solutions Architect

One order, four reactions

One order must reach billing, fulfillment, analytics, email
Direct calls to all four are brittle and slow
One event fans out to every consumer

One event fanning out from a central topic to multiple downstream services in parallel

Choosing the right messaging services

2x2 grid of AWS messaging services: SQS queue with producer-consumer flow, SNS pub-sub fanout to multiple subscribers, EventBridge event bus with rule-based routing, and Kinesis ordered shards with replay

SQS: one queue, one consumer group, at-least-once
SNS: pub/sub, push one message to many subscribers
EventBridge: route events by content to many targets
Kinesis Data Streams: ordered, replayable streaming

The fanout pattern

One event must reach several consumers at once
Publish to an SNS topic
Each SQS queue subscribes to the topic
Every consumer gets its own copy, decoupled

SNS-to-SQS fanout pattern: one SNS topic publishing to multiple SQS queues, each queue independently serving its own consumer at its own pace

Filtering so consumers get only what they need

Without filters, every subscriber gets every event
SNS subscription filter policies: filter by message attributes
EventBridge content filters: match on event fields
Each consumer receives only its relevant events

SNS subscription filter policy and EventBridge content filter: each consumer receives only events matching its filter rule, reducing unnecessary processing and cost

Routing with EventBridge

EventBridge routing diagram: event bus receiving events from multiple sources, rules matching event patterns, and matched events routing to Lambda, SQS, and other target services

Event bus: receives events from many sources
Rule: an event pattern that matches certain events
Target: where matched events are sent
One event can route to multiple targets

Streaming with Kinesis Data Streams

A continuous, real-time flow of records
- Ordered records within a shard
- Replayable: consumers re-read from any point
- Multiple consumers read the same stream independently
- Use for clickstream, telemetry, high-throughput data

Kinesis Data Streams architecture: producers write ordered records into numbered shards, three consumers independently read at different positions, with a rewind arrow showing replay capability

Lambda as an event consumer

Lambda can be triggered by SQS, SNS, EventBridge, Kinesis
Event source mapping polls SQS and Kinesis for you
SNS and EventBridge push events to Lambda
Near-real-time processing with no servers to manage

Lambda triggered by four event sources: event source mapping polling SQS and Kinesis in batches, while SNS and EventBridge push events directly to Lambda

Let's practice!

Deploying Applications on AWS