Compute capacity for Amazon Redshift Serverless - Amazon Redshift
Considerations and limitations for serverless endpoint capacityAI-driven scaling and optimization
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Compute capacity for Amazon Redshift Serverless

With Amazon Redshift Serverless you can automatically scale compute capacity up and down to match your workload requirements. Compute capacity refers to the processing power and memory allocated to your Amazon Redshift Serverless workloads. Common use cases include handling peak traffic periods, running complex analytics, or processing large volumes of data efficiently. The following terms provide details on configuring and managing compute capacity.

RPUs

Amazon Redshift Serverless measures data warehouse capacity in Redshift Processing Units (RPUs). RPUs are resources used to handle workloads.

Base capacity

This setting specifies the base data warehouse capacity Amazon Redshift uses to serve queries. Base capacity is specified in RPUs. You can set a base capacity in Redshift Processing Units (RPUs). One RPU provides 16 GB of memory. Setting higher base capacity improves query performance, especially for data processing jobs that consume a lot of resources. The default base capacity for Amazon Redshift Serverless is 128 RPUs. You can adjust the Base capacity setting from 8 RPUs to 512 RPUs in units of 8 (8,16,24...512), using the Amazon console, the UpdateWorkgroup API operation, or update-workgroup operation in the Amazon CLI.

With a minimum capacity of 8 RPU, you now have more flexibility to run simpler to more complex workloads based on performance requirements. The 8, 16, and 24 RPU base RPU capacities are targeted towards workloads that require less than 128TB of data. If your data requirements are greater than 128 TB, you must use a minimum of 32 RPU. For workloads that have tables with large number columns and higher concurrency, we recommend using 32 or more RPU.

The maximum base RPUs available, 512, adds the highest level of computing resources to your workloads. This provides more flexibility to support workloads of large complexity and accelerates loading and querying data.

Note

An expanded maximum base RPU capacity of 1024 is available in the following Amazon Web Services Regions:

  • US East (N. Virginia)

  • US East (Ohio)

  • US West (Oregon)

You can increment or decrement RPUs in units of 32 when setting a base capacity between 512-1024.

If you manage larger and more complex workloads, consider increasing the size of your Redshift Serverless data warehouse. Larger warehouses have access to more compute resources, allowing them to process queries more efficiently. Note that increasing the maximum base RPU capacity of your workgroup requires additional free IP addresses. For more information on increased free IP address requirements, go to Considerations when using Amazon Redshift Serverless.

Following are some instances where having a higher base capacity is beneficial:

  • You have complex queries that take a long time to run

  • Your tables have a large number of columns.

  • Your queries have a high number of JOINs.

  • Your queries aggregate or scan large amounts of data from an external source, such as a data lake.

For more information about Amazon Redshift Serverless quotas and limits, go to Quotas for Amazon Redshift Serverless objects.

Considerations and limitations for Amazon Redshift Serverless capacity

The following are considerations and limitations for Amazon Redshift Serverless capacity.

  • Configurations of 8 or 16 RPU support Redshift managed storage capacity of up to 128 TB. If you're using more than 128 TB of managed storage, you can't downgrade to less than 32 RPU.

  • Editing your workgroup's base capacity might cancel some of the queries running on your workgroup.

  • Amazon Redshift Serverless won't scale up your RPUs unless there are queries in the queue. Amazon Redshift Serverless won't scale up your RPUs in response to increased load from a single query. As a result, a single, resource-intensive query may cause your workgroup to run out of memory if there isn't current capacity to handle it. Ensure that your base capacity is sufficient to handle any single query that you run on your data warehouse.

AI-driven scaling and optimization

The AI-driven scaling and optimization feature is available in the all Amazon Regions where Amazon Redshift Serverless is available.

Amazon Redshift Serverless offers an advanced AI-driven scaling and optimization feature to meet diverse workload requirements. Data warehouses may have the following provisioning issues:

  • Data warehouses may be over-provisioned to improve performance of resource-intensive queries

  • Data warehouses may be under-provisioned to save costs.

Striking the right balance between performance and cost for data warehouse workloads is challenging, especially with ad-hoc queries and growing data volumes. When running mixed workloads, comprising both low and high resource-intensive queries, there is a need for intelligent scaling. The AI-driven scaling and optimization feature automatically scales Serverless compute or RPUs in response to data growth. This feature also helps maintain query performance within targeted price-performance objectives. The AI-driven scaling and optimization dynamically allocates compute resources as data volumes increase, ensuring queries continue to meet performance targets. AI-driven scaling and optimization allows the service to adapt seamlessly to changing workload requirements, without the need for manual intervention or complex capacity planning.

Amazon Redshift Serverless provides a more comprehensive and responsive scaling solution based on factors such as query complexity and data volume. This feature allows for optimizing workload price-performance while maintaining the flexibility to handle varying workloads and growing datasets efficiently. Amazon Redshift Serverless can automatically make AI-driven optimizations to your Amazon Redshift Serverless endpoint to meet your specified price-performance targets for your Serverless workgroup. This automatic price-performance optimization is especially helpful if you don't know what base capacity to set for your workloads, or if some parts of your workload might benefit from more allocated resources.

Example

If your organization typically runs workloads that only require 32 RPU but suddenly introduces a more complex query, you might not know the appropriate base capacity. Setting a higher base capacity yields better performance but also incurs higher costs, so the cost might not match your expectations. Using AI-driven scaling and resource optimization, Amazon Redshift Serverless automatically adjusts the RPUs to meet your price-performance targets while keeping costs optimized for your organization. This automatic optimization is useful regardless of workload size. The automatic optimization can help you meet your organization's price-performance targets if you have any number of complex queries.

Note

Price-performance targets are a workgroup-specific setting. Different workgroups can have different price-performance targets.

To keep costs predictable, set a limit of maximum capacity that Amazon Redshift Serverless is allowed to allocate to your workloads.

To configure price-performance targets, use the Amazon console. You must enable your price-performance target explicitly when you create your Serverless workgroup. You can also modify the price-performance target after you create the Serverless workgroup. When you enable the price-performance target, it is set to Balanced by default. .

To edit the price-performance target for your workgroup

  1. In the Amazon Redshift Serverless console, choose Workgroup configuration.

  2. Choose the workgroup for which you want to edit the price-performance target. Choose the Performance tab, then choose Edit.

  3. Choose Price-performance target, and adjust the slider to your desired setting.

  4. Choose Save changes.

  5. To update the maximum amount of RPUs that Amazon Redshift Serverless can allocate to your workload, choose the Limits tab of the Workgroup Configuration section.

You can use the Price-performance target slider to set your desired balance between cost and performance. By moving the slider, you can choose one of the following options:

  • Optimizes for cost — This setting prioritizes cost savings. Amazon Redshift Serverless attempts to automatically scale up compute capacity when doing so doesn’t incur additional charges. Amazon Redshift Serverless also attempts to scale down compute resources for lower cost, possibly increasing query runtimes.

  • Balanced — This setting creates a balance between performance and cost. Amazon Redshift Serverless scales for performance, and may result in a moderate cost increase or decrease. This is the recommended setting for most Amazon Redshift Serverless data warehouses.

  • Optimizes for performance — This setting prioritizes performance. Amazon Redshift scales aggressively for high performance, potentially incurring higher costs.

  • Intermediate positions: You can also set the slider to one of two intermediate positions between Balanced and Optimizes for cost or Optimizes for performance. Use these settings if full optimization for cost or performance is too extreme.

Considerations when choosing your price-performance target

You can use the price-performance slider to choose your desired price-performance target for your workload. The AI-driven scaling and optimization algorithm learns over time from your workload history, and improves prediction and decision accuracy.

Example

For this example, assume a query that takes seven minutes and costs $7. The following figure shows the query runtimes and cost with no scaling.

Graph for example query for Amazon Redshift Serverless autoscaling.

A given query might scale in a few different ways, as shown below. Based on the price-performance target you choose, AI-driven scaling predicts how the query trades off performance and cost, and scales it accordingly. Choosing the different slider options yields the following results:

Graph for example query for Amazon Redshift Serverless autoscaling.

  • Optimizes for Cost — With the Optimizes for Cost option, your data warehouse scales favoring choices that lower your costs. In the preceding example, the super linear scaling approach demonstrates this behavior. Scaling will only occur if it can be done in a cost-effective manner according to the scaling model predictions. If the scaling models predict that cost-optimized scaling isn’t possible for the given workload, then the data warehouse won't scale.

  • Balanced — With the Balanced option, the system scales while balancing both cost and performance considerations, with a potential limited increase in cost. The Balanced option performs superlinear, linear, and possibly sublinear workload scaling.

  • Optimizes for Performance — With the Optimizes for Performance option, in addition to the previous methods for improving performances, the system also scales even if the costs are higher, and possibly not proportional to the runtime improvement. With Optimizes for Performance, the system performs superlinear scaling, linear scaling, and sublinear scaling if possible. The closer the slider position is to the Optimizes for Performance position, the more Amazon Redshift Serverless permits sublinear scaling.

Note the following when setting the Price-Performance slider:

  • You can change the price-performance setting at any time, but the workload scaling won't change immediately. The scaling changes over time as the system learns about the current workload. We suggest monitoring a Serverless Workgroup for 1-3 days to verify the impact of the new setting.

  • The price-performance slider options Max capacity and Max RPU-hours work together. Max capacity and Max RPU-hours are the controls to limit the maximum RPUs that Amazon Redshift Serverless allows the data warehouse to scale, and the maximum RPU hours that Amazon Redshift Serverless allows the data warehouse to consume. Amazon Redshift Serverless always honors and enforces these settings, regardless of the price-performance target setting.

Monitoring resource autoscaling

You can monitor the AI-driven RPU scaling in the following ways:

  • Review the RPU capacity used graph on the Amazon Redshift console.

  • Monitor the ComputeCapacity metric under AWS/Redshift-Serverless and Workgroup in CloudWatch.

  • Query the SYS_QUERY_HISTORY view. Provide the specific query ID or query text to identify the time period. Use this time period to query the SYS_SERVERLESS_USAGE system view to find the compute_capacity value. The compute_capacity field shows the RPUs scaled during the query runtime.

Use the following example to query the SYS_QUERY_HISTORY view. Replace the sample value with your query text.

select query_id,query_text,start_time,end_time, elapsed_time/1000000.0 duration_in_seconds
from sys_query_history
where query_text like '<query_text>'
and query_text not like '%sys_query_history%'
order by start_time desc

Run the following query to see how compute_capacity scaled during the period from start_time to end_time. Replace start_time and end_time in the following query with the output of the preceding query:

select * from sys_serverless_usage
where end_time >= 'start_time'
and end_time <= DATEADD(minute,1,'end_time')
order by end_time asc

For step-by-step instructions for using these features, see Configure monitoring, limits, and alarms in Amazon Redshift Serverless to keep costs predictable .

Considerations when using AI-driven scaling and optimization

Consider the following when using AI-driven scaling and optimization:

  • For existing workloads on Amazon Redshift Serverless requiring 32 to 512 Base RPU, we recommend using Amazon Redshift Serverless AI-driven scaling and optimization for optimal results. We do not recommend using this feature for less than 32 Base RPU or more than 512 Base RPU workloads.

  • Price-performance targets automatically optimize the workload, though results may vary. We recommend using this feature over time so the system can learn your specific patterns by running a representative workload.

  • AI-driven scaling and optimization uses optimal times to apply optimizations to Serverless workgroups depending on the workload running on your Amazon Redshift Serverless instance.

To learn more about AI-driven optimizations and resource scaling, watch the following video.