Honestly, what I appreciate most about AWS SQS is how easy it is to get started and scale up. No need to mess with servers or manage any infrastructure. We rolled it out using the Serverless Framework, and the setup just clicked. Everything we needed was ready to go, no hassle.
The integrations are a huge plus, especially the way SQS hooks into AWS Lambda to trigger events. I like that the AWS Console keeps things simple. It’s easy to keep an eye on your queues, send test messages, and set up dead-letter queues. You don’t need to spend hours figuring it out. The UI is pretty straightforward.
When it comes to performance, SQS handles whatever you throw at it. Spikes in traffic? No sweat. There’s barely any lag, so it’s rock solid for separating microservices. The pricing makes sense, too. You only pay for what you use, which ends up saving us quite a bit compared to running our own message broker.
The documentation is actually helpful, and since SQS is so widely used, the community always has answers. If you hit a weird issue, someone’s probably already posted a fix. At the end of the day, SQS is basic but reliable, and the dead-letter queueing gives us just enough smarts to keep everything running smoothly. Review collected by and hosted on G2.com.
Honestly, what really gets me is how clumsy it feels handling partial redrives from a Dead Letter Queue. There’s just no straightforward way to filter and resend specific messages, like by type, attributes, or when they came in. So when a batch trips up because of some downstream bug, you’re stuck either redriving everything or messing around with your own scripts to pull, filter, and repost messages. It’s messy.
Even digging through the DLQ in the UI can be pretty frustrating. If you’re trying to track down weird, asynchronous bugs at scale, the tools just aren’t flexible enough. You spend more time wrangling the system than actually fixing the problem.
And I feel like AWS is missing out on adding some built-in smarts here. Imagine SQS could spot anomalies by itself or actually tell you what went wrong, grouping errors automatically before you even start the redrive. Right now, Lambda integration is solid, but handling partial batch failures on standard queues still means writing loads of boilerplate code.
Also, while the standard pricing is cheap, it’s easy for costs to creep up, especially if you’re running those high-frequency empty polling loops. You have to stay pretty strict with how you set up long-polling, or the bill starts sneaking up on you. Review collected by and hosted on G2.com.
