Introduction
Amazon Redshift has long been a powerhouse for cloud data warehousing, delivering high performance at a fraction of on-premises costs. With the introduction of RG instances powered by AWS Graviton, you can now achieve up to 2.2x faster performance than RA3 instances while paying 30% less per vCPU. These new instances also include an integrated data lake query engine that lets you run SQL analytics across your data warehouse and data lake from a single engine—boosting performance for Apache Iceberg by up to 2.4x and for Apache Parquet by up to 1.5x compared to RA3. Whether you're handling human-driven BI queries or high-volume agentic AI workloads, RG instances are designed to reduce costs and simplify operations. This guide walks you through the process of adopting RG instances, from assessment to migration and optimization.
What You Need
- An active AWS account with permissions to create and modify Redshift clusters.
- If migrating from an existing RA3 cluster: your cluster identifier and region, plus access to the AWS Management Console, CLI, or API.
- Understanding of your current workload patterns (query types, data volumes, concurrency) to choose the right instance size.
- (Optional) The AWS Pricing Calculator to estimate savings for your specific use case.
- Familiarity with Amazon S3 data lakes if you plan to use the integrated query engine for Iceberg or Parquet tables.
Step-by-Step Guide to Adopting RG Instances
Step 1: Assess Your Current Workload and Instance
Start by reviewing your existing Redshift setup. Identify which RA3 instance types you're using (e.g., ra3.xlplus or ra3.4xlarge) and note the vCPU, memory, and storage characteristics. The table below shows recommended RG equivalents:
| Current RA3 Instance | Recommended RG Instance | vCPU | Memory (GB) | Primary Use Case |
|---|---|---|---|---|
| ra3.xlplus | rg.xlarge | 4 | 32 | Small cluster departmental analytics |
| ra3.4xlarge | rg.4xlarge | 16 | 128 | Standard production workloads, medium data volumes |
For workloads that involve mixed warehouse and data lake queries, RG instances are particularly beneficial due to the integrated query engine.
Step 2: Estimate Cost Savings
Use the AWS Pricing Calculator to input your current RA3 instance count and expected RG instance count. Factor in the 30% lower price per vCPU and potential performance gains (up to 2.2x faster) to model total cost of ownership. Remember that faster queries can also reduce overall compute time for pay-per-query or serverless equivalents.
Step 3: Launch a New RG Cluster
If you're starting fresh, create a new cluster via the AWS Management Console, AWS CLI, or API. In the console:
- Navigate to Amazon Redshift → Clusters → Create cluster.
- Under Instance type, select the RG family (e.g.,
rg.xlarge). - Configure other settings (database name, master user, security groups) as per your requirements.
- Enable the integrated data lake query engine (it is enabled by default).
- Review and launch.
Step 4: Migrate an Existing RA3 Cluster
To migrate an existing RA3 cluster to RG:
- Take a snapshot of your current cluster (ensure it's consistent).
- Restore the snapshot to a new cluster, selecting an RG instance type during the restore process.
- Alternatively, if you prefer in-place migration, use the Modify cluster option and change the node type to the corresponding RG instance. However, this will require a brief outage; plan accordingly.
- Update your application connection strings to point to the new cluster endpoint.
- Test your queries and ETL pipelines against the new RG cluster to confirm performance improvements.
Step 5: Optimize Queries for the Integrated Data Lake Engine
With the integrated data lake query engine, you can query both warehouse tables and Amazon S3 data lakes using standard SQL. To maximize benefits:
- If you use Apache Iceberg or Parquet tables, reorganize them with partitioning and compaction best practices.
- Leverage the engine's automatic optimization for Iceberg (up to 2.4x faster) and Parquet (up to 1.5x faster).
- Monitor query performance using Amazon Redshift's Query Monitoring and Performance insights.
- For agentic AI workloads that generate high query volumes, consider using Redshift Serverless with RG instances if your volume is unpredictable—though RG is provisioned, it handles high concurrency efficiently.
Step 6: Monitor Costs and Performance
After migration, regularly review AWS Cost Explorer and Redshift metrics to validate savings. Look for reductions in price per vCPU and improvements in query latency. Set up CloudWatch alarms to notify you of unexpected spikes in usage or cost. Adjust instance sizes if workload patterns change—for example, you might need to scale up for seasonal BI peaks.
Tips and Best Practices
- Test with representative workloads: Before fully committing, run a subset of your critical queries on a small RG cluster to compare performance against RA3. This validates the 2.2x speed claim in your environment.
- Use AWS Pricing Calculator for accurate cost projections. Input your specific vCPU counts and expected query volumes to see the 30% savings.
- Plan migration during low-usage periods to minimize disruption. Snapshot-based migration is safer for production clusters.
- Combine with AWS Graviton benefits: RG instances use custom AWS Graviton chips; ensure your SQL workloads are not dependent on x86-specific features.
- Leverage the integrated query engine as a unified analytics plane. You can reduce ETL complexity by querying data directly in S3 without moving it into the warehouse.
- Monitor for agentic AI workloads: If AI agents drive a surge in queries, RG instances' higher throughput per vCPU can prevent cost explosion.
By following these steps, you can smoothly transition to Amazon Redshift RG instances and start benefiting from faster performance, lower costs, and a simplified data architecture. For more details, refer to the Amazon Redshift documentation.