Redshift RA3 vs DC2 cost: separating compute from storage

Quick answer

Redshift DC2 nodes bundle compute and local SSD storage at a fixed ratio: a dc2.large is about $0.25 per hour with 160 GB of included storage. RA3 nodes separate compute from Redshift Managed Storage, which is billed separately at about $0.024 per GB-month, and start at ra3.xlplus around $1.086 per hour. RA3 wins when your data outgrows what DC2's fixed local storage economically holds, because you scale storage without adding compute nodes. DC2 can still be cheaper for small clusters where compute matters more than storage and the data fits comfortably in local SSD.

Choosing a Redshift node type is really a choice about how compute and storage scale together. DC2 nodes couple them: each node brings a fixed amount of fast local SSD, so growing storage means adding nodes and paying for compute you may not need. RA3 nodes decouple them, letting compute and storage scale independently, with storage billed separately per GB. Which is cheaper depends almost entirely on your data-to-compute ratio.

How the two are priced

Node	Approx. hourly	Storage model
dc2.large	~$0.25	160 GB local SSD included
dc2.8xlarge	~$4.80	2.56 TB local SSD included
ra3.xlplus	~$1.086	Managed storage at ~$0.024/GB-mo
ra3.4xlarge	~$3.26	Managed storage at ~$0.024/GB-mo

With DC2 the storage is free in the sense that it is bundled, but capped: once your data exceeds the local SSD of your nodes, you must add nodes, paying for their compute purely to get storage. With RA3, compute and Redshift Managed Storage (RMS) are billed separately, so 10 TB of data costs about $240 per month in RMS regardless of how many compute nodes you run, and you size compute to the query load alone.

Where RA3 wins

RA3 is the better economic choice once storage, not compute, drives your node count. Suppose you have 8 TB of data but only need modest query performance. On DC2 you would need many dc2.large nodes just to hold 8 TB (at 160 GB each, that is 50 nodes, about $9,000 per month), almost all of it paying for unwanted compute. On RA3, a small cluster of a few ra3.xlplus nodes handles the queries while 8 TB of RMS costs about $192 per month. For any warehouse where data has grown faster than query demand, RA3 is dramatically cheaper. This is why AWS steers most new clusters to RA3.

Where DC2 can still win

DC2 remains competitive for small, compute-heavy clusters where the dataset fits comfortably in local SSD. If you have 200 GB of data and need fast queries, two dc2.large nodes at about $360 per month deliver strong performance with storage included, whereas the RA3 minimum plus RMS may cost more for that small footprint. The local SSD also gives DC2 very low-latency access. So for small data with high compute needs, DC2 can undercut RA3, though the gap has narrowed and RA3's flexibility usually wins as data grows.

Do not forget serverless and reservations

For spiky or intermittent warehouses, Redshift Serverless bills per RPU-hour and may beat any always-on cluster, as covered in Redshift Serverless cost. For steady RA3 or DC2 clusters, reserved-node pricing cuts the hourly rate substantially, the same commitment logic as reserved instances. And for query-only access to data already in S3, compare against Redshift versus Athena.

Match the node type to the ratio

The RA3-versus-DC2 decision comes down to whether storage or compute drives your cluster size. Growing datasets with moderate query load favor RA3's separated, per-GB storage; small datasets with heavy compute can still favor DC2's bundled SSD. Estimate your data size and query demand separately, and price both node types against the resource catalog before committing, so the cluster scales on the dimension that actually drives cost.

FAQ

What is the cost difference between Redshift RA3 and DC2?

DC2 bundles compute and local SSD storage: dc2.large is about $0.25 per hour with 160 GB included. RA3 separates them, billing Redshift Managed Storage at about $0.024 per GB-month on top of node cost, starting at ra3.xlplus around $1.086 per hour. RA3 lets storage scale without adding compute nodes, while DC2 caps storage at the local SSD of its nodes.

When is Redshift RA3 cheaper than DC2?

When storage rather than compute drives your node count. If you have 8 TB of data but modest query needs, DC2 would require about 50 dc2.large nodes just to hold it (around $9,000 per month), mostly paying for unwanted compute, while a small RA3 cluster plus 8 TB of managed storage costs a fraction of that. Growing datasets almost always favor RA3.

When is Redshift DC2 still worth it?

For small, compute-heavy clusters where the dataset fits comfortably in local SSD. With 200 GB of data and high query demand, two dc2.large nodes at about $360 per month deliver strong performance with storage included, and the very low-latency local SSD helps. As data grows past the local capacity, RA3's flexibility usually wins.

How is Redshift Managed Storage billed?

Redshift Managed Storage (RMS), used by RA3 nodes, is billed separately from compute at about $0.024 per GB-month in us-east-1. This decoupling means your data size sets the storage cost independent of how many compute nodes you run, so 10 TB costs about $240 per month in RMS regardless of cluster compute size.

How does C3X help with Redshift node type cost?

C3X prices Redshift clusters from Terraform including node type, node count, and storage, so the cost difference between an RA3 and a DC2 configuration is visible in the pull request. That helps teams pick the node type that matches their data-to-compute ratio before deploy, avoiding paying for compute just to hold storage on DC2 or vice versa.

What to do next

Size the cluster on the dimension that drives cost. C3X prices Redshift RA3 and DC2 from Terraform against a live resource catalog. Start with the quickstart.