Amazon Redshift is a cloud data warehouse from Amazon Web Services, run as provisioned clusters or Serverless with usage-based pricing, for teams on AWS.
Amazon Redshift runs SQL analytics over data held in its own tables and over files in an Amazon S3 data lake. Amazon Web Services builds and operates it as a managed service. It comes in two deployment options, Provisioned clusters and Redshift Serverless, and both scale to petabytes of data. The job it takes on is serving reports and analysis from one SQL engine as data volumes and query counts grow. Provisioned clusters run on RG or RA3 nodes, scale compute apart from Redshift Managed Storage, and can be paused while the data stays in place. Concurrency Scaling adds transient clusters when queries start to queue, and Automatic WLM sets a priority for each workload. Redshift Serverless provisions and scales capacity by itself, shuts down when idle, and enforces usage limits that alert or turn off queries. Redshift Spectrum queries S3 files as external tables, Apache Iceberg tables can be written to S3, and the SUPER type holds JSON. Datashares give other clusters and AWS accounts live reads, while dynamic data masking and row-level security decide what each role sees. Redshift suits data teams whose sources already live in AWS and who want their warehouse next to S3, Glue and Kinesis. Analysts and BI developers get one place to query warehouse tables and lake files together. Steady, predictable demand fits Provisioned clusters, while intermittent or spiky workloads fit Serverless, which stops when nothing runs. Organizations that spread data across many AWS accounts use datashares to give each account live access without copies. Dashboard and reporting work runs on warehouse tables, while Redshift Spectrum reaches files catalogued in the AWS Glue Data Catalog and governed through Lake Formation. Streaming ingestion pulls events from Amazon Kinesis or Amazon MSK into materialized views, which keeps operational dashboards close to live. Zero-ETL integrations bring Aurora, RDS and DynamoDB data into the warehouse without a pipeline to maintain. Redshift ML turns a SQL CREATE MODEL statement into a model trained with Amazon SageMaker, so analysts add predictions to queries they already write. It fits AWS teams that want lake, stream and application data answered by one SQL warehouse.
Pricing
View plans-
Redshift Provisioned (on demand)
Custom pricing
- $0.543 per node-hour for ra3.large, billed hourly on demand with no commitment, US East (Ohio)
- RG nodes from $0.3801 per hour (rg.large) to $9.128 per hour (rg.12xlarge)
- RA3 nodes from $0.543 per hour (ra3.large) to $13.04 per hour (ra3.16xlarge)
- Previous-generation DC2 nodes at $0.25 (dc2.large) and $4.80 (dc2.8xlarge) per hour
- Redshift Managed Storage on RG and RA3 billed apart at $0.024 per GB-month
- Pausing a cluster suspends on-demand billing; storage is still charged
- One hour of Concurrency Scaling credit per cluster per day, up to 30 hours banked
- Concurrency Scaling past the credit at the cluster's per-second rate, one-minute minimum
- Reserved Instances on 1 or 3 year terms (No, Partial or All Upfront) at lower rates than on demand
- Two-month trial of up to 750 hours a month in Regions without Serverless
-
Redshift Serverless
Custom pricing
- $0.36 per RPU-hour, billed per second for capacity used, no commitment, US East (Ohio)
- Each RPU carries 16 GB of memory; base capacity set from 4 to 1024 RPUs
- Capacity scales automatically and shuts down when idle
- Redshift Managed Storage billed apart at $0.024 per GB-month
- Max RPU-hours and MaxRPU settings cap compute use
- Serverless Reservations cut compute cost by up to 24% on a 1-year term or up to 45% on a 3-year term
- $300 trial credit for compute and storage, valid 90 days, for accounts new to Serverless
-
Redshift Spectrum
Custom pricing
- $5.00 per TB of data scanned, rounded up to the next megabyte, 10 MB minimum per query
- Billed on top of the cluster or Serverless compute that runs the query
-
Redshift ML
Custom pricing
- $20 per million cells for the first 10M cells, $15 for the next 90M, $7 above 100M
- Two CREATE MODEL requests a month for two months under the SageMaker free tier, up to 100,000 cells each
- Small Amazon S3 charges for the data CREATE MODEL generates
Features
- Included: Independent scaling of storage and compute
- Included: Serverless query execution
- Included: Multi-cluster concurrency scaling
- Not included: Time travel to prior table states
- Not included: Zero-copy cloning of databases
- Included: Auto-refreshing materialized views
- Included: Native semi-structured columns
- Included: External tables on data lake files
- Included: Open table format interoperability
- Included: Streaming row ingestion into tables
- Included: Live data sharing across accounts
- Included: Column masking policies
- Included: Row access policies
- Included: Automatic query result reuse
- Included: Workload management queues
- Included: Compute spend limits and alerts
- Not included: Cross-region warehouse replication
- Not included: In-warehouse Python execution
Additional Features
- Redshift Provisioned clusters on RG and RA3 nodes
- Redshift Serverless workgroups
- Redshift Managed Storage
- Pause and resume for provisioned clusters
- Concurrency Scaling transient clusters
- Redshift Spectrum queries on Amazon S3 files
- Apache Iceberg table writes to S3 and S3 table buckets
- SUPER data type with PartiQL navigation
Best for
- Teams whose data already sits in Amazon S3 and other AWS services
- Analysts and BI developers running SQL over large datasets
- Data engineers streaming events from Kinesis or Amazon MSK
- Steady, predictable workloads suited to provisioned clusters
- Intermittent or spiky workloads suited to Serverless
- Organizations sharing live data across many AWS accounts
- Teams enforcing row and column security inside the warehouse
- Data lake teams standardizing on Apache Iceberg tables
Use cases
- Serving BI dashboards from a SQL data warehouse
- Querying Parquet and JSON files in S3 in place
- Feeding near-real-time dashboards from Kinesis streams
- Bringing Aurora, RDS or DynamoDB data in through zero-ETL
- Sharing live tables with another AWS account
- Masking sensitive columns per user or role
- Training ML models with a SQL CREATE MODEL statement
- Querying nested JSON with PartiQL
- Absorbing query spikes with Concurrency Scaling
- Capping Serverless compute with usage limits
- Writing Apache Iceberg tables to S3 table buckets
- Running petabyte-scale analytics on AWS
Screenshots & Videos
Explore Amazon Redshift in action
User Reviews
Compare with Similar Software
Click to compare side by side with Amazon Redshift
