3 AWS event services. 3 completely different guarantees.
It’s easy to pick SQS, SNS, or EventBridge based on what you used last. Then production hits. You discover the service you picked can’t do ordering, can’t fan out, can’t replay, or can’t filter the way you need.
Start with 5 properties your event requires:
<ol>
<li>Delivery model: point-to-point or fan-out?</li>
<li>Ordering: does order matter?</li>
<li>Filtering: do consumers need different subsets?</li>
<li>Replay: do you need to re-process events later?</li>
<li>Schema awareness: is this a cross-team contract?</li>
</ol>
Each AWS service was built for a different combination of these.
🔹 SQS: THE WORK QUEUE
Use when: one producer, one consumer, and you need a buffer.
SQS is a pull-based queue. Consumers poll. Messages stay until processed or they hit the DLQ. Standard SQS gives you at-least-once delivery. FIFO SQS gives you exactly-once processing and strict ordering, but caps you at lower msg/sec per message group.
Real use case: Stripe uses SQS-style queues to buffer webhook retries. If your downstream service is slow, the queue absorbs the spike.
Avoid SQS when you need multiple independent consumers of the same event. You would have to build fan-out manually.
🔹 SNS: THE FAN-OUT BROADCASTER
Use when: one event, many consumers, and you want push delivery.
SNS is pub/sub. You publish once, and every subscriber gets a copy. Subscribers can be SQS queues, Lambdas, or HTTP endpoints. The common production pattern is SNS to multiple SQS queues. This gives you fan-out plus per-consumer buffering and retry.
SNS has message filtering. But it has no replay. Once a message is delivered, it is gone forever. Ordering only exists in FIFO SNS, which inherits the same 300 msg/sec throttle.
Avoid SNS when you need event replay, schema validation, or complex routing rules.
🔹 EVENTBRIDGE: THE EVENT ROUTER
Use when: cross-service routing, SaaS integrations, or scheduled events.
EventBridge is built for event-driven architectures at scale. You get a schema registry, content-based filtering with full JSON path matching, archive-and-replay built in, and native SaaS integrations.
Real use case: Bamboo Health processes 250+ million healthcare events per day through EventBridge. They use archive-and-replay to re-run events after code fixes. That is the superpower no other AWS event service has.
The tradeoff: EventBridge has higher latency (usually under 500ms, but 50% slower than SNS) and costs ~$1 per million events vs $0.50 for SNS. For high-volume, low-latency paths, SNS still wins.
🔹 THE DECISION CHEAT SHEET
• One producer to one consumer buffering? -&gt; SQS
• One event to many consumers with low latency? -&gt; SNS to SQS fan-out
• Cross-service routing, replay, or schema contracts? -&gt; EventBridge
• Strict ordering at &lt; 300 msg/sec? -&gt; SQS FIFO or SNS FIFO
• Scheduled cron events? -&gt; EventBridge
hashtag#SystemDesign hashtag#AWS hashtag#DistributedSystems hashtag#Microservices hashtag#BackendEngineering

You can share you thought with this squad 

SQS VS SNS VS EventBridge