This year I finally got the chance to attend AWS re:invent in person, and I have enjoyed the experience a great deal. Here are the launches that I found most exciting.

  1. The most exciting new release for me is Aurora DSQL, a Postgres compatible multi-region strongly consistent database.
    1. DSQL relies on Amazon Time Sync and the clock-bound library to produce reliable timestamps for comparison
    2. DSQL uses optimistic concurrency control at commit time to minimize the impact cross region coordination
    3. This product is still in preview. Although it is postgres compatible, it’s missing many common features in its current state (doc)
  2. Apache Iceberg related features:
    1. S3 Table bucket for Iceberg
      1. This new S3 bucket type provides great performance improvement at about 10% higher storage cost than regular S3 bucket. It also provides Iceberg maintenace feature like compactino. The maintenance logic for now is similar to AWS Glue optimization.
    2. S3 Metadata provides queryable metadata for S3 objects using Iceberg on S3 Tables.
    3. AWS Firehose can now replicate data from Aurora source to Iceberg and supports schema evolution.
    4. AWS Glue now supports zero-ETL into Iceberg from sources like DynamoDB and Salesforce.
  3. I did not spend much time on AI features during this re:invent, but the one I found most interesting is Amazon Bedrock Guardrails, because it uses automated reasoning to help prevent hallucinations from GenAI.