dataonaws
— Stack · Data

Amazon Redshift

A columnar cloud data warehouse. You load data once into managed storage and query it many times, and two ways of running it decide how you pay.

How it fits

Data lands once in managed storage. Everything else queries it.

Sourcesdatabases, S3, streamszero-ETLCOPYAMAZON REDSHIFTQuery planningleader / workgroupComputenodes or RPUsManaged storagecolumnar, billed per GB-month apart from computeSpectrum · data sharing · concurrency scalingBI and dashboardsdbt, SQL jobsS3 data lakeread in place with Spectrummetrics + query history → CloudWatch, system tables
Two ways to run it

Same SQL, two billing models.

Provisioned

RA3 clusters

Nodes you size and pay per hour, from about $0.543 an hour. Fits steady load; reserved instances lower the price.

Serverless

RPUs

Capacity in RPUs, billed per second while queries run, 60 seconds minimum. From 4 RPUs, about $1.50 an hour.

Either way

Managed storage

Billed per GB-month, separately from compute, so storage grows without adding nodes.

Prices for US East (N. Virginia), check the current price for your region.

The three choices that matter

Most of the bill and the speed comes from here.

01

How much base capacity

Serverless starts from the base capacity you set, and the default is 128 RPUs. Small and medium workloads often run fine much lower.

02

How rows are spread

Keys on the columns you join and filter on keep work on each node. Check them before adding capacity.

03

Where the data sits

Hot tables in managed storage, history read in place from S3, other teams reading through data sharing instead of a copy.

* On provisioned clusters, 10 MB minimum per query. On Serverless, Spectrum queries are billed as RPU time.

What you pay

A Redshift bill, in one line.

Compute
node-hours or RPU-seconds
The largest line on most bills
+
Storage
GB-month
Managed storage, separate from compute
+
Lake queries
TB scanned
Spectrum on provisioned; included in RPU time on Serverless
+
Bursts
concurrency scaling
Free credits first, then per second
What we see in production
Serverless left at the default base capacity.Lowering it is often the first saving on small workloads.
Provisioned clusters running nights and weekends.Pause them, or move irregular workloads to Serverless.
Joins that move data between nodes.Revisit distribution keys before buying more capacity.
How we help

Build & optimize your data platform on AWS

State of the art, costs under control. Most bills grow with habits, not with data.

Read the approach →
PDF · coming soon

The one-page Amazon Redshift field guide.

Best practices, cost traps and performance checks, in the order you should run them. Printable, no fluff.

One email when the PDF is ready. Nothing else unless you tick the box.