Quick answer:
A Redshift cluster combines a leader node with one or more compute nodes that run queries in parallel. Create one in the Redshift console under Clusters > Create cluster, choosing RG or RA3 nodes. New clusters are private and encrypted by default.
You can edit, reboot, resize, pause, snapshot and delete a running cluster. Elastic resize takes about ten minutes on average; pausing stops on-demand compute billing. In US East (N. Virginia), the smallest node, rg.large, costs $0.3801 per node-hour. Redshift Serverless skips cluster management entirely.
Amazon Redshift is a cloud service from Amazon Web Services, which provides a fully-managed analytical data warehouse. One of Redshift’s key features is its clustering mechanism, which enables it to handle Petabyte-scale data at high speed.
Cluster Basics
A cluster is the core unit of operations in the Amazon Redshift data warehouse. Each Redshift cluster is composed of two main components:
- Compute Node, which has its own dedicated CPU and memory, determined by the node type. Compute nodes execute queries and you can have many nodes in one cluster. On RG and RA3 nodes, the data itself lives in Redshift Managed Storage, which combines local SSDs with Amazon S3, so you scale and pay for compute and storage separately.
- Leader Node, which manages communication between the compute nodes and the client applications. The leader node compiles code, distributes the compiled code to the compute nodes, and assigns a portion of the data to each compute node.
Redshift’s cluster and node organization represents a Massively Parallel Processing (MPP) architecture. MPP is a technology that allows multiple processors to perform many operations at a time.
Some key advantages of MPP architecture for databases are:
- Ability to query large volumes of data at speed.
- Good for analytic workloads, which require sophisticated queries.
- Linearly scalable to cater for data growth.
- Flexible enough to incorporate unstructured and semi-structured data.
Provisioned Clusters vs. Redshift Serverless
Redshift Serverless replaces cluster management with a workgroup for compute and a namespace for databases and their data. Capacity is measured in Redshift Processing Units (RPUs). In US East (N. Virginia), compute costs $0.375 per RPU-hour, billed per second with a 60-second minimum; base capacity can be as low as 4 RPUs, or $1.50 per hour. You pay for compute only while it runs.
AWS recommends reserved nodes for steady-state production workloads, where they reduce the hourly rate. Provisioned clusters also suit fixed node sizing, Multi-AZ on a fixed footprint, or development and test clusters paused on a schedule. This page covers provisioned clusters; the Redshift architecture chapter compares both approaches in more depth.
Cluster Management Options
There are four main options for managing clusters in Redshift.
- Console: The console is the main dashboard on Redshift that lets you manage your data. In the console, you can create, modify, and delete clusters by clicking a few buttons.
- AWS SDKs: You can perform cluster management operations from your own code with Amazon’s software development kits. You can interact with Redshift using an SDK for any one of several platforms, including Java, .NET, PHP, Python, Ruby, and Node.js.
- AWS CLI: The AWS command line interface lets you manage your clusters with command line operations, using the aws redshift commands. You can run cluster operations from your preferred terminal program.
- Amazon Redshift API: You can also use the Amazon Redshift management API to manage clusters. To use this method, you call the API by submitting a HTTP or HTTPS request, using GET or POST with a parameter named Action.
Direct API calls leave low-level details such as error handling and request signing to your application, making an SDK or the CLI usually simpler. Resize, pause and resume operations can also run on a schedule configured in the console or with the CLI and API.
The remainder of this article gives you a quick overview on how to create a cluster in the Redshift Console and perform common cluster operations.
Creating a Cluster
Parameters
In Redshift, every cluster is associated with a parameter group. The parameter group contains the settings used to configure the database. If you fail to choose a parameter group, Redshift allocates its own default parameter group (default.redshift-2.0) to your cluster. The default group can’t be modified, so to change a value you create a custom parameter group.
In the console, you can create or modify parameter groups under Configurations > Workload management: choose a parameter group, open the Parameters tab and select Edit parameters.
The main cluster parameters are:
- auto_analyze Turns automatic analyze on or off. ANALYZE is an operation used to update table statistics, and with auto_analyze on (the default), Redshift runs it in the background.
- datestyle Use this parameter to specify the format of data (ISO, Postgres, SQL, or German) and year/month/day ordering (DMY, MDY, YMD). The default is ISO, MDY.
- enable_user_activity_logging Lets you choose whether to enable Redshift to create a log of information about connections and user activities in your database.
- extra_float_digits Sets the number of digits displayed for floating-point values.
- query_group A character string that you use to apply a user-defined label to a group of queries.
- require_ssl Choose whether your cluster should require an SSL connection to allow client connections. Since January 10, 2025, the default is true.
- search_path This parameter is a comma-separated list of existing schema names. The parameter specifies the order in which schemas are searched when a table or other object is referenced with no schema component.
- statement_timeout Any statement taking longer than your specified number of milliseconds gets aborted. The default, 0, sets no limit.
- wlm_json_configuration Lets you define the number of query queues that are available, and how queries are routed to those queues for processing. The default is automatic WLM, which lets Redshift manage concurrency and memory itself.
After changing parameter values, reboot every associated cluster to apply them; parameter status shows pending-reboot until then. Dynamic WLM properties apply without a reboot; static WLM properties require one. A SET command overrides a parameter for your current session only.
Main Steps
The main steps for creating a cluster in the Redshift Console are:
- Sign in to the AWS Management Console and open the Amazon Redshift console. Choose Clusters in the navigation menu, then click the Create cluster button.
- In the Cluster configuration section, input the cluster’s basic information:
- Cluster identifier: A unique name for your cluster using 1 to 63 lowercase alphanumeric characters or hyphens. It must start with a letter and can’t end with a hyphen or contain two consecutive hyphens.
- In the same section, choose Help me choose to get a sizing suggestion based on your data size and query characteristics, or I’ll choose to pick the Node type and number of Nodes yourself. You can also select Load sample data to load the TICKIT sample dataset into the dev database.
Node Type
Choose one of the two current node families:
- RG: Graviton-based nodes, generally available since May 2026, which AWS recommends for the best price performance. RG clusters include an integrated data lake query engine that runs on the cluster’s own compute.
- RA3: The earlier managed-storage nodes, which query data lakes through Redshift Spectrum.
Both families use Redshift Managed Storage, billed separately at $0.024 per GB-month. DC2 and DS2 node types are no longer available.
The table shows US East (N. Virginia) on-demand list prices.
| Node type | vCPU | Memory | Managed storage per node | Nodes per cluster | US East on-demand price per node-hour |
|---|---|---|---|---|---|
| rg.large | 2 | 16 GiB | 1 TB (single node) or 8 TB | 1 or 2 to 16 | $0.3801 |
| rg.xlarge | 4 | 32 GiB | 32 TB | 2 to 16 | $0.7602 |
| rg.4xlarge | 16 | 128 GiB | 128 TB | 2 to 32 | $3.04267 |
| rg.12xlarge | 48 | 384 GiB | 128 TB | 2 to 128 | $9.128 |
| ra3.large | 2 | 16 GiB | 1 TB (single node) or 8 TB | 1 or 2 to 16 | $0.543 |
| ra3.xlplus | 4 | 32 GiB | 4 TB (single node) or 32 TB | 1 or 2 to 16 | $1.086 |
| ra3.4xlarge | 12 | 96 GiB | 128 TB | 2 to 32 | $3.26 |
| ra3.16xlarge | 48 | 384 GiB | 128 TB | 2 to 128 | $13.04 |
Number of Nodes
Select the appropriate number of nodes to store your data. You can choose a single-node cluster, which contains one node that shares leader and compute functionality, or a multi-node cluster, which has a leader node and a user-specified number of compute nodes. The leader node is offered free-of-charge by AWS for a multi-node cluster. Single-node clusters are possible only on rg.large, ra3.large and ra3.xlplus, and AWS does not recommend them for production workloads.
- In the Database configuration section, set the admin credentials:
- Admin user name: An account name for the admin user of the cluster (older versions of the console called it the master user).
- Admin password: Generate a password, add one manually, or choose Manage admin credentials in AWS Secrets Manager, which stores and rotates the password for an additional Secrets Manager fee.
- Optionally, add IAM roles under Cluster permissions so the cluster can load data from Amazon S3 and other AWS services. Then, under Additional configurations, turn off Use defaults if you want to change these settings:
- Database name: An optional field for creating a custom database name. If you leave it blank, the database is named dev.
- Database port: A port number used for connecting to the database from client applications. The default is 5439; on RG and RA3 clusters the port must be between 5431 and 5455 or between 8191 and 8215.
- Cluster parameter group lets you associate a parameter group with the cluster.
- Encryption: New clusters are encrypted by default with an AWS-owned key; you can choose your own AWS KMS key instead.
- Network and security lets you choose the VPC, the cluster subnet group, VPC security groups, enhanced VPC routing and whether the cluster is publicly accessible (off by default).
- Maintenance, Monitoring and Backup let you pick a maintenance track, add a CloudWatch alarm on disk usage, and set automated snapshot retention (one day by default) and cluster relocation.
- Click Create cluster. The cluster might take several minutes to be ready to use; when its status shows Available, you can connect to it.
To run your first queries, select the cluster and choose Query data, then Query in query editor v2, a web-based SQL client that needs no download.
Further reading:
Common Cluster Operations
The main cluster operations are:
- Modify: Used to make changes to the cluster. In today’s console this is the Edit button.
- Delete: Used to remove a cluster you no longer need.
- Reboot: Restart the cluster so that changes requiring a reboot can take effect.
- Resize: Change the node type, cluster type, or number of nodes.
- Pause and resume: Suspend on-demand compute billing while you don’t need the cluster, then bring it back.
Modify
To modify a cluster, go through these steps:
- Open the Redshift console and select Clusters. Select the cluster you want to modify.
- Choose Edit to open the Edit cluster page.
- Make your desired changes and choose Save changes. Common changes include setting a new admin password, changing the cluster identifier, and modifying the cluster parameter group. Further reading: Modifying a cluster
Delete
Follow these steps to delete a cluster:
- Open the Redshift Console, select Clusters, and select the cluster you want to delete.
- Choose Actions, then Delete.
- On the Delete cluster page, choose whether to create a final snapshot of the cluster, type delete to confirm, and choose Delete cluster.
If you delete a cluster without creating a snapshot, all data gets deleted permanently. The delete process typically takes several minutes. Create a snapshot if you plan to use the same data and configuration again.
Deleting a cluster removes its automated snapshots. Manual snapshots, including the final snapshot, survive cluster deletion and incur backup storage charges.
Further reading: Deleting a Cluster
Reboot
To reboot a cluster:
- Select Clusters in the Redshift console, and select the cluster you want to reboot.
- Choose Actions, then Reboot cluster.
- Confirm the reboot. The cluster status shows rebooting until the cluster is available again.
The main reason to restart a cluster by rebooting it is to apply changes to a parameter group: any pending cluster modifications are applied at this reboot. Note that you can’t access the cluster while it reboots.
Further reading: Rebooting a Cluster
Resize
To resize a cluster:
- Select Clusters in the Redshift console, and select the cluster you want to resize.
- Choose Actions, then Resize.
- On the Resize cluster page, choose the new node type and number of nodes, and whether to resize now, once at a specific time, or on a recurring schedule.
- Choose Resize now or Schedule resize, then wait for the resize to finish.
The ability to scale your data warehouse with little interruption is one of Redshift’s most important features, but a resize is not free of downtime. Redshift offers two types of resize:
- Elastic resize: AWS’s recommended first choice takes about ten minutes on average, with a few minutes of unavailability; most queries pause and connections stay open, though some queries can drop. Changing node type keeps data read only during copying, then drops connections when the endpoint switches.
- Classic resize: Use it for changes outside elastic resize limits. The cluster restarts, closes all connections and becomes unavailable for a few minutes; reads and writes then continue during migration if prerequisites are met, including a manual snapshot no more than 10 hours old.
Neither resize can be cancelled once it starts, and changing node type does not change an existing reserved-node reservation.
Further reading: Resizing a Cluster
Pause and Resume
Choose Actions, then Pause, to pause immediately, at a set time or on a schedule. Pausing stops on-demand compute billing, leaving storage charges. It typically completes within 15 minutes; resuming takes several minutes, with some queries slower while data re-hydrates.
Resume a paused cluster before modifying it. Secrets Manager secrets continue to incur charges while paused. Create a final snapshot before deleting a paused cluster if you need its data; without one, you can’t restore it.
Snapshot
Snapshots are point-in-time backups of a cluster. You can take a snapshot manually, or have Redshift create snapshots automatically. Snapshots can be copied to another AWS region for resilience purposes.
Automated snapshots default to about every eight hours or every 5 GB per node of data changes, whichever comes first. Default retention is one day, configurable from 1 to 35 days on RG and RA3. Manual snapshots remain until deletion or retention expiry. Custom schedules allow at most one snapshot per hour.
It’s possible to restore an entire cluster or a table from a historic snapshot. Below we show how to restore a table.
To restore a table from a snapshot using the Amazon Redshift console:
- Sign in to the Amazon Redshift console.
- Select Clusters and choose the cluster that holds the table.
- Choose Actions, then Restore table.
- Choose the snapshot, then enter the source database, schema and table, and a new name for the target table.
- Click Restore table to restore the table.
Restore one table at a time into the running source cluster, using a new target table name.
For further reading, and to learn about additional cluster snapshot options, see Working with Snapshots.
Creating a Cluster in a VPC
An Amazon Virtual Private Cloud (VPC) is a cloud service that lets you launch Amazon Redshift and other AWS resources in a virtual network.
The two main benefits of an Amazon VPC are:
- Clusters in a VPC are private by default: the default security group only allows inbound traffic from resources in the same group, and you open access deliberately with security group rules, public access or Redshift-managed VPC endpoints.
- Enhanced VPC routing, which lets you tightly manage the flow of data between your Amazon Redshift cluster and all of your data sources by forcing COPY and UNLOAD traffic through your VPC.
Enhanced VPC routing defaults to off, routing COPY and UNLOAD traffic through the internet. Changing it restarts the cluster; enabling it requires an S3 gateway endpoint, NAT gateway or internet gateway.
Setting up Redshift Clusters in a VPC
Every RG and RA3 cluster runs in a VPC: AWS retired the older EC2-Classic platform in 2023, so launching outside a VPC is no longer an option. Setting up a cluster in a VPC is similar to how you set one up in the console, with a few additional steps.
- You first need to set up a VPC. You can either choose the default VPC in your account or one that you create. Dedicated-tenancy VPCs are not supported.
- Create an Amazon Redshift cluster subnet group specifying which VPC subnets can be used by the Amazon Redshift cluster.
- Give access to inbound connections for a VPC security group that will use the cluster.
- Set up the cluster using the console as previously detailed. When you get to the Additional configurations section, turn off Use defaults and input the details of your VPC under Network and security.
Choose the subnet group carefully: it can’t be changed after cluster creation. Keep network ACL and route-table settings consistent across its subnets because nodes can move during resize. A group spanning several Availability Zones enables RG and RA3 cluster relocation, on by default, or a Multi-AZ deployment. Multi-AZ runs compute in two Availability Zones behind one endpoint at the same per-node hourly rates.
Wrapping Up
Amazon Redshift is a fully managed cloud-based data warehouse in which you provision clusters of cloud-based computing resources (named nodes). You can manage a cluster using the Redshift console, the AWS SDKs, the AWS CLI, or by using the Amazon Redshift management API.
To create a cluster, you choose a cluster parameter group, which holds the settings used to configure the database. You then input the main cluster details such as cluster identifier, admin user name and admin password, before choosing a node type and number of nodes for your cluster.
RG nodes, built on AWS Graviton, are the current recommendation, while RA3 nodes remain available; both keep data in Redshift Managed Storage, so compute and storage scale separately. Clusters on Redshift are either single-node or multi-node.
Additional configuration options available during cluster setup include encryption, which is on by default, and network settings. Every cluster now launches inside a Virtual Private Cloud (VPC), so the main network decisions are the subnet group, security groups, public access and enhanced VPC routing. A VPC lets you launch Amazon Redshift in a virtual network for added security and greater control over data pipelines.
Once a cluster is launched, you can modify, delete, reboot, resize, or pause that cluster. The resize operation is an important Redshift feature that lets you scale your data warehouse with only a few minutes of unavailability. You can also create cluster snapshots, which are point-in-time cluster backups that you can restore the cluster to in the event of a crash.
Amazon Redshift’s clustering mechanism uses Massively Parallel Processing (MPP) and columnar storage to analyze huge volumes of data at high speed. Redshift is ideal for enterprises that need a truly cloud-based representation of an on-premise data warehouse without the high entry costs, significant logistical concerns, and scalability challenges associated with on-site data warehouses (see our data warehouse architecture and data warehouse concepts traditional vs. cloud comparison).
To fill a new cluster, many teams load data with AWS Glue or another ELT service and model it inside the warehouse with dbt. If cluster management is more work than you want, Snowflake, BigQuery and Databricks are the usual alternatives to compare in the directory.
Common questions
How long does it take to resize a Redshift cluster?
Elastic resize averages about ten minutes, including a few minutes of unavailability. Classic resize restarts the cluster, then permits reads and writes during migration when prerequisites are met, including a manual snapshot no more than 10 hours old. Neither resize can be cancelled once started.
How much does a Redshift cluster cost?
US East (N. Virginia) on-demand compute costs $0.3801 to $13.04 per node-hour, plus $0.024 per GB-month of managed storage. Multi-node clusters have a free leader node. Total cost depends on Region, node type, count and reservations. Compare reserved-node savings for steady workloads using the warehouse pricing playbook.
Can I pause a Redshift cluster?
Yes. Choose Actions, then Pause, to pause immediately or on a schedule. Pausing typically completes within 15 minutes and stops on-demand compute billing. Storage and Secrets Manager charges continue. Resume before modifying the cluster, and create a final snapshot before deleting it if you need its data.
Should I use a provisioned cluster or Redshift Serverless?
Serverless suits spiky or unpredictable workloads and teams avoiding node sizing, at $0.375 per RPU-hour in US East (N. Virginia), billed per second with a 60-second minimum. Provisioned clusters suit steady workloads where reserved nodes cut hourly rates, or teams wanting fixed sizing or a Multi-AZ deployment.
Where to next
Read Redshift Architecture for how nodes, slices and managed storage fit together, and Redshift Columnar Storage 101 for storage mechanics. Then explore BigQuery architecture and Redshift vs. BigQuery. For the basics, see the difference between a database and a data warehouse and Data Mart vs. Data Warehouse.
Snowflake, BigQuery, Redshift, Databricks, Fivetran, Airbyte and the rest, with pricing and what each one replaces.
Browse the tools