Modernize Amazon Redshift: RA3 to RG Migration best practices

This article provides best practices and guidance for migrating Amazon Redshift clusters from RA3 to Graviton-based RG instances, which offer up to 2.2x faster performance and 30% lower costs.

• RG instances deliver 2.2x faster data warehouse and 2.4x faster data lake workloads compared to RA3 at 30% lower price per vCPU
• Three migration approaches: Elastic Resize (recommended, 10-15 minutes), Classic Resize (supports all configurations), and Snapshot/Restore (allows validation without affecting source)
• RG.xlarge maps 1:1 to RA3.xlplus; RG.4xlarge requires 3 nodes per 4 RA3.4xlarge nodes for clusters with 3-64 nodes
• Elastic Resize maintains cluster endpoint and minimal downtime; Classic Resize rebalances data but takes longer; Snapshot/Restore enables parallel setup
• Zero-ETL tables temporarily resync during resize; data sharing unavailable briefly; AWS DMS CDC tasks resume automatically after cluster availability
• Key best practices: notify downstream teams, schedule during maintenance windows, take manual snapshots, test target configuration, and verify downstream applications post-migration

RG instances maintain feature parity with RA3 while delivering superior performance and cost efficiency, making migration recommended for all RA3 customers.