Simon Willison’s Weblog

Introducing Amazon Aurora DSQL (via) New, weird-shaped database from AWS. It's (loosely) PostgreSQL compatible, claims "virtually unlimited scale" and can be set up as a single-region cluster or as a multi-region setup that somehow supports concurrent reads and writes across all regions. I'm hoping they publish technical details on how that works at some point in the future (update: they did), right now they just say this:

When you create a multi-Region cluster, Aurora DSQL creates another cluster in a different Region and links them together. Adding linked Regions makes sure that all changes from committed transactions are replicated to the other linked Regions. Each linked cluster has a Regional endpoint, and Aurora DSQL synchronously replicates writes across Regions, enabling strongly consistent reads and writes from any linked cluster.

Here's the list of unsupported PostgreSQL features - most notably views, triggers, sequences, foreign keys and extensions. A single transaction can also modify only up to 10,000 rows.

No pricing information yet (it's in a free preview) but it looks like this one may be true scale-to-zero, unlike some of their other recent "serverless" products - Amazon Aurora Serverless v2 has a baseline charge no matter how heavily you are using it. (Update: apparently that changed on 20th November 2024 when they introduced an option to automatically pause a v2 serverless instance, which then "takes less than 15 seconds to resume".)