Skip to main content
In this lesson you’ll learn what AWS Redshift Serverless is, which problems it solves, how capacity is measured and billed, and the configuration knobs you use to balance performance and cost. Redshift Serverless removes the need to provision and run a persistent cluster. Instead of paying for fixed CPU, memory, and storage 24/7, Serverless automatically scales compute to match workload demand so you only pay for the capacity you actually use.
A slide titled "Why do we need Redshift Serverless?" showing a graph of serverless capacity with variable spikes above a 24 RPU baseline and a note of "Max: 600 hours," plus bullet points explaining that Redshift Serverless automatically scales data warehouse capacity (measured in RPUs) and you pay per RPU-hour.

​
How Redshift Serverless is measured and billed

Redshift Serverless measures compute capacity in Redshift Processing Units (RPUs). Billing is based on RPU-hours (charged on a per-second basis) only while your data warehouse is active. This billing includes queries that access data in open file formats on S3. Two primary configuration values control capacity:
  • Base capacity — the guaranteed RPU capacity kept available to serve queries.
  • Max capacity — an upper cap you set to limit the RPU usage and contain costs; policies determine what happens when usage reaches the cap.
Now let’s break down RPUs and sizing considerations.
A slide titled "Redshift Serverless RPUs" with five colorful rounded cards summarizing capacity rules. They state: Base Capacity; 1 RPU = 16 GB RAM; adjustable between 8 and 512 RPUs; 8–24 RPUs support up to 128 TB; and a minimum of 32 RPUs is required for more than 128 TB.

​
RPU facts and sizing guidance

Key points:
  • Increasing base capacity gives more memory and compute headroom, improving performance for resource-intensive queries (large joins, wide tables, or many concurrent users).
  • Base capacity should be chosen to balance cost and expected workload — pick higher RPUs for steady, heavy workloads and lower for intermittent/ad hoc queries.
Choose base capacity to match your workload pattern: use higher RPUs for wide tables, many concurrent users, or heavy ETL/aggregation jobs; use lower RPUs for sporadic, light queries to reduce cost.
For catalogs with very wide tables or consistently high concurrency, AWS recommends selecting 32 or more RPUs to ensure adequate memory and compute capacity.

​
Benefits of Redshift Serverless

A presentation slide titled "Redshift Serverless – Features" with two colored panels. The left orange panel reads "Cost Saving" with a coin and down-arrow icon, and the right purple panel reads "Optimal Cluster Usage" with a cluster/network icon.
  • Cost efficiency: Pay only for compute when the service is active — avoids the always-on cost of provisioned clusters.
  • Reduced operational overhead: No manual provisioning or cluster management; automatic scaling handles workload variability.
  • Simplified sizing: The need to guess node types and counts is reduced because Serverless scales capacity automatically.

​
Summary and operational considerations

A presentation slide titled "Summary Redshift Serverless" listing three brief points: that Redshift Serverless is an on-demand data warehouse that auto-manages scaling and provisioning, RPUs measure computational capacity and cost, and RPUs automatically scale with workload.
  • Redshift Serverless is an on-demand data warehouse that auto-manages provisioning and scaling for you.
  • RPUs are the unit of computational capacity and directly drive performance and billing.
  • RPU consumption scales automatically with workload so queries are executed efficiently without manual intervention.
A presentation slide titled "Summary Redshift Serverless" with a turquoise gradient panel on the left and three numbered points (05–07) on the right summarizing RPU capacity, storage limits, and cost forecasting. The right side lists details about adjusting RPUs, managed storage up to 128 TB, and RPU usage for cost forecasting.
Operational reminders and best practices:
  • Remember: 1 RPU = 16 GB memory. Adjust base capacity in supported increments (commonly 8 RPUs).
  • Managed storage up to 128 TB is supported when base capacity is in the 8–24 RPUs range; use at least 32 RPUs if you need more than 128 TB.
  • Monitor RPU usage metrics and set alerts to forecast costs. Configure max capacity to cap spending and apply policies that control behavior at the cap.
  • Use workload testing (representative queries and concurrency) to determine the appropriate base capacity before going to production.

​
When to choose Serverless vs. provisioned Redshift

  • Choose Serverless when you want minimal cluster management, variable query traffic, or pay-for-what-you-use billing.
  • Choose provisioned clusters when you require consistent, high-performance throughput with predictable sizing, or when you need fine-grained control over node types and distribution styles.
With these concepts, you can evaluate whether Redshift Serverless fits your workload and cost goals, and select base/max RPU settings that match your performance and storage requirements.

Watch Video