Amazon Redshift RG instances now include rg.large and rg.12xlarge, so I tried launching an rg.large

Amazon Redshift RG instances now include rg.large and rg.12xlarge, so I tried launching an rg.large

Amazon Redshift's Graviton-based RG instances now include rg.large and rg.12xlarge, expanding the family to four sizes. I launched rg.large in the Tokyo Region and checked how it works.
2026.07.17

This page has been translated by machine translation. View original

Introduction

On July 16, 2026, two new sizes, rg.large and rg.12xlarge, were added to the Amazon Redshift RG instances.

https://aws.amazon.com/jp/about-aws/whats-new/2026/07/amazon-redshift-adds-rg-large-12xlarge-instance-sizes/

The architecture of RG instances and the unified data lake engine are explained in the previous article.

https://dev.classmethod.jp/articles/amazon-redshift-rg-instances-graviton/

RG Instance Specification List

With this addition, the RG family now consists of the following four sizes.

Instance vCPU Memory Slices/Node RMS Limit/Node Node Range
rg.large (newly added) 2 16 GiB 2 8 TB 2–16
rg.xlarge 4 32 GiB 2 32 TB 2–16 (resize: 32)
rg.4xlarge 16 128 GiB 8 128 TB 2–32 (resize: 64)
rg.12xlarge (newly added) 48 384 GiB 16 128 TB 2–128

Source: Amazon Redshift Management Guide — Clusters and nodes

※ As of July 16, 2026, rg.large / rg.12xlarge are not yet reflected in the specification table of the Management Guide. The specifications of the new sizes are compiled based on the product attributes of the Pricing API and the options available when creating a cluster. Of these, the RMS limit and the number of slices for rg.large were also verified in an actual environment.

rg.large has 2 vCPU / 16 GiB and is the smallest size in the RG family. It is the most affordable entry point for benefiting from RG (Graviton + unified data lake engine). rg.12xlarge has 48 vCPU / 384 GiB and is the largest size in the RG family.

Launching rg.large

I created a 2-node rg.large cluster in the Tokyo region.

aws redshift create-cluster \
  --region ap-northeast-1 \
  --cluster-identifier rg-large-test \
  --cluster-type multi-node \
  --node-type rg.large \
  --number-of-nodes 2 \
  --master-username admin \
  --master-user-password '****'

After the cluster became available, the main parameters confirmed with describe-clusters are as follows.

Item Value
Parameter group default.redshift-2.0
Encryption Enabled by default (AWS_OWNED_KMS_KEY)
AQUA disabled
Node configuration LEADER + COMPUTE-0 + COMPUTE-1
Elastic Resize target 3 or 4 nodes
MultiAZ Disabled
MaintenanceTrack current

Note that rg.large and rg.12xlarge are available only on the current track (P202).

Checking System Information via DB Connection

I ran SQL with the Redshift Data API (aws redshift-data execute-statement / get-statement-result) and checked the results.

version()

PostgreSQL 8.0.2 on i686-pc-linux-gnu, compiled by GCC gcc (GCC) 3.4.2 20041017 (Red Hat 3.4.2-6.fc3), Redshift 1.0.358853

stv_node_storage_capacity

node used (MB) capacity (MB)
0 47 8,000,000
1 22 8,000,000

capacity = 8,000,000 MB (8 TB). This matched the RMS limit in the specification table (8 TB/node).

stv_slices

node slice localslice type
0 0 0 D
0 1 1 D
1 2 0 D
1 3 1 D

2 nodes × 2 slices/node = 4 slices in total. This was as per the specification table (2 slices/node).

Checking Pricing with the Pricing API

By specifying instanceType and regionCode with aws pricing get-products, you can retrieve the on-demand price for each size and region.

aws pricing get-products \
  --region us-east-1 \
  --service-code AmazonRedshift \
  --filters \
    'Type=TERM_MATCH,Field=instanceType,Value=rg.large' \
    'Type=TERM_MATCH,Field=regionCode,Value=ap-northeast-1' \
    'Type=TERM_MATCH,Field=productFamily,Value=Compute Instance'

I compiled the on-demand prices for all four RG sizes. The following are the unit prices per compute node.

Instance vCPU Memory Tokyo (ap-northeast-1) US East (us-east-1)
rg.large (newly added) 2 16 GiB $0.4473/h $0.3801/h
rg.xlarge 4 32 GiB $0.8946/h $0.7602/h
rg.4xlarge 16 128 GiB $3.58027/h $3.04267/h
rg.12xlarge (newly added) 48 384 GiB $10.7429/h $9.128/h

Prices were retrieved from the Pricing API as of 2026-07-16. For the latest prices, please check Amazon Redshift Pricing.

Cleanup

Since the verification was complete, I deleted the cluster.

aws redshift delete-cluster \
  --region ap-northeast-1 \
  --cluster-identifier rg-large-test \
  --skip-final-cluster-snapshot

Summary

RG instances have been expanded to four sizes, and you can now choose from rg.large with 2 vCPU to rg.12xlarge with 48 vCPU. This makes it easier to choose a size suited to your workload, from small-scale verification use cases to large-scale configurations.

The on-demand price of rg.large in the Tokyo region is $0.4473 per node-hour, and the compute cost of the minimum 2-node configuration is $0.8946 per hour. As the smallest size in the RG family, it serves as an entry point for trying out the RG architecture.

https://dev.classmethod.jp/articles/amazon-redshift-rg-instances-graviton/

https://dev.classmethod.jp/articles/20260517-amazon-redshift-rg-vs-ra3/

https://dev.classmethod.jp/articles/20260517-amazon-redshift-rg-migration/

https://dev.classmethod.jp/articles/20260516-amazon-redshift-benchmark-data/

https://dev.classmethod.jp/articles/20260626-amazon-redshift-ri-upfront-pricing-rg-instances/

https://dev.classmethod.jp/articles/20260709-amazon-redshift-rg-instances-trailing-track/

https://dev.classmethod.jp/articles/redshift-rg-instances-track-switch/

Official Documentation


コスト最適化、打ちっぱなしで元通りになっていませんか

タグ付けも不要リソースの棚卸しも、施策は打てる。でも続ける仕組みがなければ、コストは数か月でじわじわ戻る。一度きりで終わらせず、FinOpsを組織に定着させる=CCoEの役割。最適化を回し続ける進め方を、無料資料にまとめました。

CCoE総合支援

FinOpsを定着させる資料をもらう

Share this article

AWSのお困り事はクラスメソッドへ