Your multi-tenant SQS queue has a problem you don’t see coming.
One tenant sends 10,000 messages. Another sends 10.
Both share the same queue. Guess whose messages sit there for hours?
This is the noisy neighbor problem. It has been hitting multi-tenant SQS architectures silently for years.
AWS just fixed it.
FAIR QUEUES FOR SQS
Launched this week, here is what actually happens under the hood.
SQS monitors in-flight messages per tenant in real time.
When one tenant has a disproportionately large number of in-flight messages, SQS identifies it as a noisy neighbor. It automatically prioritizes delivery for the quiet tenants.
The noisy tenant still gets processed. They just cannot monopolize the queue anymore.
HOW TO ENABLE IT
Add a MessageGroupId to your outgoing messages:
SendMessageRequest request = new SendMessageRequest()
.withQueueUrl(queueUrl)
.withMessageBody(messageBody)
.withMessageGroupId(“tenant-123”);
That is it. Zero consumer code changes required, and no API latency impact.
THE KEY METRIC TO WATCH: DWELL TIME
Dwell time is the total time a message spends in the queue from arrival to processing.
Without Fair Queues: one noisy tenant spikes, and everyone’s dwell time rises.
With Fair Queues: quiet tenants maintain low dwell times even during traffic surges.
AWS added a new CloudWatch metric (ApproximateNumberOfMessagesVisibleInQuietGroups) so you can confirm your quiet tenants are actually protected.
WHY NOT JUST USE FIFO QUEUES?
FIFO queues solve noisy neighbors by limiting in-flight messages per tenant. But that kills throughput for high-volume tenants.
Fair Queues gives you fairness without the throughput cap.
WHO SHOULD CARE?
• Multi-tenant SaaS with shared SQS queues
• Event-driven systems where one producer can spike
• Architectures where proportional processing matters more than strict ordering
Before this, you had two bad options. Separate queues per tenant (expensive and complex), or accept that big customers crush small ones.
Now you have a better option — let SQS handle it.
#AWS #SystemDesign #DistributedSystems #SoftwareEngineering #Microservices

You can share you thought with this squad 

Backend Developer 

AWS launched Fair Queues for SQS, addressing the noisy neighbor problem in multi-tenant queue architectures. By adding a MessageGroupId to outgoing messages, SQS now monitors in-flight messages per tenant in real time and automatically prioritizes delivery for quiet tenants when one tenant dominates. No consumer code changes are needed. A new CloudWatch metric (ApproximateNumberOfMessagesVisibleInQuietGroups) lets you verify quiet tenants are protected. Unlike FIFO queues, Fair Queues provide fairness without capping throughput for high-volume tenants, eliminating the need for separate per-tenant queues.