![[Update] Amazon Redshift rg.large instances now support single-node clusters](https://images.ctfassets.net/ct0aopd36mqt/wp-thumbnail-63f1274931b942e9a92e601c1127ad73/cfe87ec6d62fa2fc3c474ed4cb2f6c2e/amazon-redshift?w=3840&fm=webp)
[Update] Amazon Redshift rg.large instances now support single-node clusters
This page has been translated by machine translation. View original
This is Ishikawa from the Cloud Business Division. The Amazon Redshift rg.large instance featuring the AWS Graviton processor now supports single-node clusters.
Previously, RG instances required a configuration of 2 or more nodes, but now you can create a cluster with just 1 node, making it possible to set up validation environments and PoCs (proof of concept) that don't require high availability at a lower cost.
What is the Amazon Redshift RG Instance?
This is a node type for Amazon Redshift provisioned clusters based on the AWS Graviton processor, which became generally available in May 2026. Compared to the previous generation RA3 instances, it processes data warehouse and data lake workloads up to 2.4 times faster, with a 30% lower price per vCPU.
A key feature of RG instances is the built-in vectorized data lake query engine independently developed by Amazon Redshift. Since data from Apache Iceberg and Parquet is processed directly on the cluster's compute nodes, SQL analytics spanning both the data warehouse and data lake can be executed with a single engine.
Update Details
The main changes are as follows.
- You can now create a single-node (1 node) cluster with rg.large instances
- Single-node rg.large is available in patch versions P204 and later
- A cost-efficient option has been added for small workloads that do not require high availability
- PoCs and tests can be launched in a short time
RG Instance Node Type List
With the addition of single-node support, two specifications are defined for rg.large: one for single-node and one for multi-node.
| Node Type | vCPU | Memory (GiB) | Managed Storage Limit per Node | Node Count Range |
|---|---|---|---|---|
| rg.large (single-node) | 2 | 16 | 1 TB | 1 |
| rg.large (multi-node) | 2 | 16 | 8 TB | 2–16 |
| rg.xlarge (multi-node) | 4 | 32 | 32 TB | 2–16 |
| rg.4xlarge | 16 | 128 | 128 TB | 2–32 |
| rg.12xlarge | 48 | 384 | 128 TB | 2–128 |
Note that the managed storage limit for single-node is 1 TB, which is smaller than the 8 TB available in multi-node configuration.
Differences Between Single-Node and Multi-Node Configurations
When Amazon Redshift runs with 2 or more compute nodes, it mirrors the data on each node to the disk of another node to reduce the risk of data loss. Since this mirroring does not function in a single-node configuration, it is not suitable for production workloads requiring high availability.
Differences in Data Lake Query Execution Location
In RA3 provisioned clusters, queries to the data lake are executed on the Redshift Spectrum fleet, which is independent of the cluster. In RG clusters, on the other hand, data lake queries are processed using the cluster's own compute resources.
With RG clusters, the restriction that existed when using Redshift Spectrum with RA3—"only Amazon S3 buckets in the same region"—has been removed, and queries can now be executed against Amazon S3 buckets in other regions as well. However, cross-region queries incur additional data transfer charges.
Supported Regions
Single-node rg.large is available in the following regions.
- US East (N. Virginia, Ohio)
- US West (N. California, Oregon)
- Africa (Cape Town)
- Asia Pacific (Hong Kong, Tokyo, Seoul, Osaka, Mumbai, Hyderabad, Singapore, Sydney, Jakarta, Melbourne, Malaysia, Taipei, Thailand)
- Canada (Central)
- Europe (Frankfurt, Stockholm, Spain, Ireland, London, Paris)
- South America (São Paulo)
- Mexico (Central)
- AWS GovCloud (US-East, US-West)
Both the Tokyo Region and Osaka Region are included.
Impact on Pricing
RG instances are priced 30% lower per vCPU compared to RA3 instances. Additionally, since RG clusters process data lake queries using the engine within the cluster, the per-scan charges for Redshift Spectrum ($5 per TB scanned) do not apply.
However, while the Redshift Spectrum scan charges are eliminated, the cluster's own operating costs and data transfer charges for cross-region access still apply separately. Please estimate the total cost across your entire workload. For the latest pricing, please refer to the Amazon Redshift pricing page.
How to Use
You can create a single-node cluster from the AWS Management Console using the following steps.
- Sign in to the AWS Management Console and open the Amazon Redshift console (
https://console.aws.amazon.com/redshiftv2/) - Select "Clusters" from the navigation menu
- Select "Create cluster"
- In the cluster settings, choose "I'll choose" and specify
rg.largefor "Node type" and1for "Number of nodes" (specifying1for "Number of nodes" will not result in an error) - Enter the database settings, network and security settings, and select "Create cluster"
If you are unsure about sizing, selecting "Help me choose" in step 4 will launch a sizing calculator tool.
Usage Notes
When adopting a single-node configuration, please consider the following points.
- The managed storage limit is 1 TB per node. Since this is smaller than the 8 TB available in multi-node configuration, you need to estimate your data volume
- Since there is only one compute node, data mirroring is not performed. Please choose a multi-node configuration for workloads requiring high availability. If operating with a single node, we recommend confirming your acceptable RTO/RPO and the recovery procedure from snapshots in advance
- The patch version requirements differ between new creation and migration from an existing cluster. When migrating from an RA3 cluster to a single-node rg.large, the source cluster's patch version must be P204 or later (migrations from DC2 clusters have no patch version requirement)
- Workloads that heavily use Redshift Spectrum may experience degraded query performance after migrating to RG. In that case, consider increasing the number of nodes or migrating to a larger RG node type
Conclusion
The Amazon Redshift rg.large instance now supports single-node clusters, making it possible to use RG instances starting from just 1 node. While you should be aware that the managed storage limit becomes 1 TB and data mirroring is not performed, this is a sufficiently practical configuration for validation environments that do not require high availability.
If you have wanted to try RG instances but found that a multi-node configuration was not cost-effective, or if you want to evaluate the performance of the vectorized data lake query engine for Apache Iceberg and Parquet, why not start a PoC with a single-node rg.large cluster?
Related Articles
