SQS vs SNS: Differences and Which the AWS Exam Expects (2026)

Illustration comparing a line of envelopes entering one box with a glowing orb sending beams to several receivers

Amazon SQS is a queue: messages wait until one consumer pulls and processes them, for up to 14 days. Amazon SNS is publish/subscribe: each message is pushed straight away to every subscriber, and SNS doesn’t store it for later. On the exam, choose SQS to buffer work and decouple a producer from a consumer, SNS to notify many systems at once, and both together (SNS fanning out to several SQS queues) when many systems each need a reliable copy.

Facts verified November 2026 against AWS documentation.

SQS vs SNS at a glance

In short: choose SQS to queue work for one consumer, SNS to push one message to many subscribers, and both together to fan out reliably.

Amazon SQSAmazon SNS
ModelMessage queuePublish/subscribe (topics)
DeliveryPull: consumers poll the queuePush: SNS sends to every subscriber
Who gets each messageOne consumer processes it, then deletes itEvery subscriber gets a copy
StorageMessages kept 4 days by default, configurable from 60 seconds to 14 daysNot stored. SNS retries delivery, and failed messages can go to a dead-letter queue
Subscribers / consumersYour application, AWS Lambda, other services that pollAmazon SQS, AWS Lambda, HTTP/S endpoints, email, SMS, mobile push, Amazon Data Firehose
Ordering and duplicatesStandard: at-least-once, best-effort order. FIFO: ordered within a message group, with deduplicationStandard topics, or FIFO topics that deliver in order to SQS queues
FilteringConsumers get every message in the queueSubscription filter policies send each subscriber only the messages it wants
Main exam purposeDecoupling, buffering, absorbing spikesFan-out, notifications, alerts

Choose SQS when…

  • A producer is faster than the consumer, and you need a buffer so nothing is lost during spikes.
  • Work must be processed once by one worker, for example orders, image-resize jobs or batch tasks.
  • The consumer may be offline for a while. Messages wait in the queue for up to 14 days.
  • You need strict order and no duplicates. That’s an SQS FIFO queue.
  • You want to scale workers on the backlog. Auto Scaling can follow the queue depth.

Choose SNS when…

  • One event must reach several systems at once, for example “order placed” goes to billing, shipping and analytics.
  • People need a notification by email, SMS or mobile push.
  • A CloudWatch alarm must alert someone. Alarms publish to an SNS topic.
  • Subscribers only want some messages. SNS filter policies route them without extra code.

Why use SNS and SQS together?

The fan-out pattern combines both: you publish once to an SNS topic, and SNS delivers a copy to several SQS queues, one per consuming system. Each system then reads its own queue at its own pace. This fan-out pattern gives you SNS’s one-to-many delivery and SQS’s durability: if the shipping system is down for an hour, its messages wait in its queue instead of being lost.

On the exam, fan-out is the answer whenever a question says one event must be processed by several independent applications and none of them may miss a message. For strict ordering end to end, use an SNS FIFO topic with SQS FIFO queues.

Exam cues: which answer the question wants

Cue in the questionLikely answerExam
“Decouple components”, “buffer requests”, “absorb traffic spikes”Amazon SQSCLF-C02, SAA-C03
“Send notifications by email or SMS”, “publish/subscribe”Amazon SNSCLF-C02, SAA-C03
“Process in the exact order received”, “no duplicates”SQS FIFO queueSAA-C03
“One event, several applications, each must process every message”SNS topic fanning out to multiple SQS queuesSAA-C03
“Subscribers only receive relevant messages”SNS subscription filter policiesSAA-C03
“Messages that fail repeatedly must be kept for analysis”Dead-letter queueSAA-C03, SOA-C03
“Email the operations team when a CloudWatch alarm fires”CloudWatch alarm → SNS topicSOA-C03, SAA-C03
“Scale workers on the number of messages waiting”SQS + Auto Scaling on queue depthSAA-C03, SOA-C03
“Route events from SaaS apps or AWS services by content rules”Amazon EventBridge (not SNS)SAA-C03
“Real-time streaming, replay, multiple consumers reading the same stream”Amazon Kinesis Data Streams (not SQS)SAA-C03

Worked example 1: the order spike

An online shop’s web tier sends orders straight to a processing application on EC2. During sales, orders arrive faster than they can be processed, and some are lost. The company wants no lost orders and the ability to scale processing on demand. What should a solutions architect do?

The problem is a fast producer and a slow consumer that are tightly coupled. What’s missing is a buffer that holds the work until it can be done. SNS would push every order to the processing application immediately, which doesn’t help when that application is overloaded, and SNS doesn’t keep messages for later. An SQS queue between the web tier and the processors stores every order durably until a worker pulls it. Then an Auto Scaling group can add processing instances as the number of messages in the queue grows. If the question added “each order must be processed exactly once and in order”, the answer would be an SQS FIFO queue.

Worked example 2: one order, three systems

When an order is placed, the billing, shipping and analytics systems must each process it. They run at different speeds, and none of them may miss an order, even if one is temporarily unavailable. Which design meets these requirements?

Three systems need the same message, so a single SQS queue won’t work: one consumer would take each message and the others would never see it. SNS alone delivers to all three, but if the shipping system is down, it can miss messages after SNS’s retries run out. The design that meets both requirements is fan-out: publish the order to one SNS topic and subscribe three SQS queues, one per system. Each system reads its own queue at its own pace, and nothing is lost while a system is down.

Common SQS vs SNS traps

  1. “SNS stores messages.” It doesn’t. If you need durability, put an SQS queue behind the topic.
  2. One SQS queue for several consumer types. Each message is processed by only one consumer. Use a queue per system with SNS fan-out.
  3. Standard queues for ordered work. Standard SQS gives best-effort ordering and at-least-once delivery. “Exactly once, in order” needs FIFO.
  4. SQS vs Kinesis. SQS deletes a message once it’s processed. If several applications must read the same stream of records, or you need to replay data, the exam wants Kinesis Data Streams.
  5. SNS vs EventBridge. Both deliver events to many targets, but EventBridge is the answer when the question stresses content-based routing rules, SaaS event sources or a schedule.
  6. The visibility timeout. If processing takes longer than the visibility timeout, the message becomes visible again and another worker processes it twice. Raise the timeout to fit the processing time.

SQS vs SNS FAQ

Test yourself

Decoupling questions are a staple of the Solutions Architect exam.

Free mock exam

Take SAA-C03 Mock Exam 1, free

A full-length, timed mock in the real exam format, with every answer explained so you know exactly where you stand.

65

questions

130

minutes

720

to pass

Free account required.

Then practise full scenarios with our SAA-C03 practice exams.

New to AWS? Warm up with our free AWS Cloud Practitioner practice exam.

Read the SAA-C03 exam guide for the full blueprint, or compare containers in ECS vs EKS.


AWSBoy is not affiliated with Amazon Web Services. AWS and related marks are trademarks of Amazon.com, Inc. or its affiliates.

Sources (checked November 2026): Amazon SQS message quotas · Amazon SQS FIFO queues · Fanout to Amazon SQS queues · SNS message delivery for FIFO topics · Amazon SQS features · Amazon SNS features

Shopping Cart
Scroll to Top