Has anyone here tried using Aurora DSQL in their production systems? I'm curious about how it's working out for you and whether it's beginner-friendly. Have you faced any challenges while adopting it or transitioning from another system?
5 Answers
I liked working with DSQL, but it was quite a hassle with migrations. Surprisingly, some DDL features aren’t supported yet, which forced me to tweak my migration tool quite a bit to make it work. I still struggled with being fully reversible.
Right now, I’m moving two production apps to DSQL. Honestly, I've found it to be significantly slower than my previous system, particularly with complex queries – sometimes up to 10 times slower, which was quite a shock.
That sounds tough! If you're comfortable, it'd be great to discuss your performance experiences more in-depth. This seems really surprising.
I can't help but feel that Aurora DSQL may not be the best fit for primary production use cases when compared to Aurora Postgres. It seems better suited for certain auxiliary tasks, rather than being primary.
I would disagree! The active/active capabilities are why we switched over, and it works well for our needs.
I've worked with Aurora DSQL a lot, and it's pretty impressive! It's a distributed SQL database that's compatible with Postgres, but it's definitely not the same as Aurora Postgres. I chatted with the product team recently, and they've got some exciting features lined up for early 2026 that are already in beta, so that's a good sign! The learning curve isn't too steep if you're familiar with Postgres, you just need to adjust your mindset to treat it like its own system.
That's promising! I'm looking for a change stream feature similar to what DynamoDB Streams offers. Without it, I’m hesitant to transition.
Is there any chance the roadmap with those new features will be shared publicly?
What about pricing compared to DynamoDB for less frequently used applications? I'm curious how the DPU-based pricing stacks up.
I'm also unsure about that; with DynamoDB, it's pretty straightforward to estimate costs. It'll be interesting to see how AWS formulates the pricing for DSQL.
For key-value apps, I think the pricing will be close enough. For SQL-heavy workloads, it'll depend on how you use it, but I’d expect it to be similar to Aurora Postgres on TPC-C workloads.
I've heard that more DDL support, including the ability to drop columns, is in the pipeline. What other improvements are you hoping for?