Amazon Redshift's new RG instances: 2.2x faster than RA3 with zero data lake scan fees!

Hey there, it's Shii-chan! Today I've got a treat for everyone who loves data analytics: Amazon Redshift just got a brand-new family member!

AWS Blog aws.amazon.com

What was announced?

Over on the AWS Blog, Amazon has announced new AWS Graviton-based RG instances for Amazon Redshift. And it's not just about raw speed: an engine that queries your S3 data lake directly is built right in from the start!

RG instances run data warehouse workloads up to 2.2x as fast as the older RA3 instances, and they cost 30% less per vCPU.

The story so far

Redshift has been evolving since 2013, moving from early dense-compute nodes to RA3 instances and serverless. When you wanted to query a data lake (files sitting in S3), you'd reach for Amazon Redshift Spectrum, which charged $5 per TB scanned. As query volumes balloon, especially with AI agents firing off requests, that per-terabyte bill really starts to add up.

What changes

With RG instances, data lake queries run from the very same engine. On top of that...

• Apache Iceberg tables run up to 2.4x as fast as RA3, and Apache Parquet up to 1.5x
• Data lake queries stay inside your own VPC and reuse your existing IAM roles
• That $5-per-TB scanning charge? Gone, down to zero!

So we move from "treat the warehouse and the data lake as separate things" to "one engine that handles both, faster and cheaper."

Dive Deep

Amazon Redshift RG instances are built on AWS Graviton. Here's where you'd move from your current RA3:

There are two migration paths. For compatible configurations, Elastic Resize takes only 10-15 minutes of downtime; if you want to change your configuration, use Snapshot and Restore. Your external tables, schemas, and existing Amazon Redshift Spectrum queries keep working unchanged, so there's nothing to worry about.

Check the numbers on the Amazon Redshift pricing page, and find the finer migration details in the Redshift Management Guide.

Availability is broad: Asia Pacific (including Tokyo, Osaka, Seoul, and Singapore), plus the US, Canada, Europe, and South America.

Wrap-up

• Redshift now has AWS Graviton-based RG instances
• Data warehouse workloads run up to 2.2x faster than RA3 at 30% lower cost per vCPU
• The data lake query engine is integrated, dropping the $5-per-TB scanning charge to zero
• Iceberg up to 2.4x, Parquet up to 1.5x, all inside your VPC with your existing IAM roles
• Migrate via Elastic Resize or Snapshot and Restore; existing Spectrum queries still work

If you run a sizable Redshift cluster and you've started worrying about both query volume and cost, this update is going to land perfectly for you!