AWS athena documentation change
Summary
Updated documentation for Athena capacity reservations with clarified terminology, added details about query quotas, queue behavior, DPU allocation mechanics, and operational considerations. Expanded examples for isolation and sharing use cases.
Security assessment
Changes focus on functional aspects of capacity management (quota impacts, queue times, billing details) without addressing authentication, authorization, vulnerability mitigation, or data protection. No security vulnerabilities or explicit security controls are mentioned in the diff.
Diff
diff --git a/athena/latest/ug/capacity-management.md b/athena/latest/ug/capacity-management.md index 22343a831..ce036e376 100644 --- a//athena/latest/ug/capacity-management.md +++ b//athena/latest/ug/capacity-management.md @@ -9 +9 @@ Understand DPUsConsiderations and limitations -You can use capacity reservations to get dedicated processing capacity for the queries you run in Athena. With capacity reservations, you can take advantage of workload management capabilities that help you prioritize, control, and scale your most important interactive workloads. For example, you can add capacity at any time to increase the number of queries you can run concurrently, control which workloads can use the capacity, and share capacity among workloads. Capacity is fully-managed by Athena and held for you as long as you require. Setup is easy and no changes to your SQL statements are required. +You can use capacity reservations to get dedicated serverless processing capacity for the queries you run in Athena. With capacity reservations, you can take advantage of workload management capabilities that help you prioritize, control, and scale your most important workloads. For example, you can add capacity to control the number of queries you can run at the same time, choose which workloads can use the capacity, and share capacity among workloads. Capacity is serverless and fully-managed by Athena and held for you as long as you need it. Setup is easy and no changes to your SQL queries are required. @@ -13 +13 @@ To get processing capacity for your queries, you create a capacity reservation, -Workgroups play an important role when you use capacity reservations. Workgroups allow you to organize queries into logical groupings. With capacity reservations, you selectively assign capacity to workgroups so that you control how the queries for each workgroup behave and how they are billed. For more information about workgroups, see [Use workgroups to control query access and costs](./workgroups-manage-queries-control-costs.html). +Workgroups play an important role when you use capacity reservations. Workgroups allow you to organize queries into logical groupings or use cases. With capacity reservations, you selectively assign capacity to workgroups so that you control how the queries for each workgroup behave and how they are billed. For more information about workgroups, see [Use workgroups to control query access and costs](./workgroups-manage-queries-control-costs.html). @@ -15 +15 @@ Workgroups play an important role when you use capacity reservations. Workgroups -Assigning workgroups to reservations lets you give priority to the queries that you submit to the assigned workgroups. For example, you could allocate capacity to a workgroup used for time-sensitive financial reporting queries to isolate those queries from less critical queries in another workgroup. This enables consistent query execution for critical workloads while allowing other workloads to run independently. +Assigning workgroups to capacity reservations lets you give priority to these queries because they run on your reserved capacity and do not count towards your DDL and DML query quota. For example, you can allocate capacity to a workgroup used for time-sensitive financial reporting queries to isolate those queries from less critical queries in another workgroup. This gives you predictable query execution for critical workloads while allowing other workloads to run independently. @@ -19 +19 @@ You can use capacity reservations and workgroups together to meet different requ - * Isolation – To isolate an important workload, you assign a single workgroup to one reservation. Only queries from the assigned workgroup use the processing capacity from the chosen reservation. + * Isolate important queries – To ensure that an important workload has the capacity it needs when you need it, create a capacity reservation and assign its workgroup to the reservation. Only queries from the assigned workgroup use the processing capacity from your reservation. For example, to ensure reliable execution of queries that support a production application, assign the production workgroup for those queries to a capacity reservation. When developing queries, use a separate workgroup that is not associated with a reservation and move the queries to the production workgroup when ready. @@ -21,3 +21 @@ You can use capacity reservations and workgroups together to meet different requ - * Sharing – Multiple workloads can share capacity from one reservation. For example, if you want a predictable monthly cost for a specific set of workloads, you can assign multiple workgroups to a single reservation. The assigned workgroups share the reservation's capacity. - - * Mixed model – You can use capacity reservations and per-query billing at the same time in the same account. For example, to ensure reliable execution of queries that support a production application, you assign a workgroup for those queries to a capacity reservation. When developing the queries before you move them to the production workgroup, you use a separate workgroup that is not associated with a reservation and therefore uses per-query billing. + * Share capacity across similar workloads – Multiple workloads can share capacity from one reservation. This allows you to achieve a predictable cost for these workloads and control their concurrency. For example, if you have scheduled workloads that are tolerant to delayed query execution start times, you can assign their workgroups to a single reservation. This frees up your DDL and DML query quota for interactive queries that run in the same account, ensuring these queries start with minimal delay. @@ -30,3 +28 @@ You can use capacity reservations and workgroups together to meet different requ -Capacity is measured in Data Processing Units (DPUs). DPUs represent the compute and memory resources used by Athena to access and process data on your behalf. One DPU provides 4 vCPUs and 16 GB of memory. The number of DPUs that you specify influences the number of queries that you can run concurrently. For example, a reservation with 256 DPUs can support approximately twice the number of concurrent queries than a reservation with 128 DPUs. - -You can create up to 100 capacity reservations with up to 1,000 total DPUs per account and region. The minimum number of DPUs that you can request is 24. If you require more than 1,000 DPUs for your use case, please reach out to [[email protected]](mailto:[email protected]?subject=Athena Provisioned Capacity DPU Limit Request). +Capacity is measured in Data Processing Units (DPUs). DPUs represent the serverless compute and memory resources used by Athena to access and process data on your behalf. One DPU typically provides 4 vCPUs and 16 GB of memory. The number of DPUs that you hold influences the number of queries that you can run concurrently. For example, a reservation with 256 DPUs can support approximately twice the number of concurrent queries than a reservation with 128 DPUs. @@ -38 +34 @@ For information about estimating your capacity requirements, see [Determine capa - * The feature requires [Athena engine version 3](./engine-versions-changing.html). + * You can use capacity reservations and per-query billing, based on data scanned, at the same time in the same account. @@ -40 +36 @@ For information about estimating your capacity requirements, see [Determine capa - * A single workgroup can be assigned to at most one reservation at a time, and you can add a maximum of 20 workgroups to a reservation. + * Queries run on capacity reservations do not count towards your DDL and DML query quota. @@ -42 +38 @@ For information about estimating your capacity requirements, see [Determine capa - * You cannot add Spark enabled workgroups to a capacity reservation. + * If your capacity is busy serving other queries, newly submitted queries are queued until capacity is available. The maximum allowed time in queue is 10 hours. @@ -44 +40 @@ For information about estimating your capacity requirements, see [Determine capa - * To delete a workgroup that has been assigned to a reservation, remove the workgroup from the reservation first. + * A workgroup can be assigned to one capacity reservation at a time. You can assign a total of 20 workgroups to a single reservation. When you assign multiple workgroups to a reservation, capacity is shared across workgroups and allocated to queries based on their submission order. There may be variation in execution order due to how Athena dynamically allocates capacity to queries. @@ -46 +42 @@ For information about estimating your capacity requirements, see [Determine capa - * The minimum number of DPUs you can provision is 24. + * Athena automatically allocates between 4 and 124 DPUs to DML queries based on their complexity. DDL queries consume 4 DPUs each. Refer to the following topics for more information: @@ -48 +44 @@ For information about estimating your capacity requirements, see [Determine capa - * You can create up to 100 capacity reservations with up to 1,000 total DPUs per account and region. + * [Determine capacity requirements](./capacity-management-requirements.html) @@ -50 +46 @@ For information about estimating your capacity requirements, see [Determine capa - * Requests for capacity are not guaranteed and can take up to 30 minutes to complete. + * [Control capacity usage](./capacity-management-control-capacity-usage.html) @@ -52 +48 @@ For information about estimating your capacity requirements, see [Determine capa - * There is a minimum billing period of 1 hour per reservation. After 1 hour, capacity is billed per minute. For pricing information, see [Amazon Athena pricing](https://aws.amazon.com/athena/pricing/). + * The minimum number of DPUs required with each capacity reservation is 24. There is a minimum billing period of 1 hour for the initial allocation of capacity to a reservation. After 1 hour, capacity is billed per minute. For pricing information, see [Amazon Athena pricing](https://aws.amazon.com/athena/pricing/). @@ -54 +50 @@ For information about estimating your capacity requirements, see [Determine capa - * Reserved capacity is not transferable to another capacity reservation, AWS account, or AWS Region. + * You can create up to 100 capacity reservations with up to 1,000 total DPUs per account and region. If you require more than 1,000 DPUs for your use case, please reach out to [[email protected]](mailto:[email protected]?subject=Athena Provisioned Capacity DPU Limit Request). @@ -56 +52 @@ For information about estimating your capacity requirements, see [Determine capa - * DDL queries on capacity reservations consume DPUs. + * Requests for capacity are not guaranteed and can take up to 30 minutes to complete. Capacity is not transferable to another capacity reservation, AWS account, or AWS Region. @@ -58 +54 @@ For information about estimating your capacity requirements, see [Determine capa - * Queries that run on provisioned capacity do not count against your active query limits for DDL and DML. + * The `DPUConsumed` CloudWatch metric is per-workgroup rather than per-reservation. Thus, if you move a workgroup from one reservation to another, the `DPUConsumed` metric includes data from the time when the workgroup belonged to the first reservation. For more information about using CloudWatch metrics in Athena, see [Monitor Athena query metrics with CloudWatch](./query-metrics-viewing.html). @@ -60 +56 @@ For information about estimating your capacity requirements, see [Determine capa - * If all DPUs are in use, submitted queries are queued. Such queries are not rejected and do not go to on-demand capacity. + * To delete a workgroup that has been assigned to a reservation, remove the workgroup from the reservation first. @@ -62 +58 @@ For information about estimating your capacity requirements, see [Determine capa - * The `DPUConsumed` CloudWatch metric is per-workgroup rather than per-reservation. Thus, if you move a workgroup from one reservation to another, the `DPUConsumed` metric includes data from the time when the workgroup belonged to the first reservation. For more information about using CloudWatch metrics in Athena, see [Monitor Athena query metrics with CloudWatch](./query-metrics-viewing.html). + * Workgroups configured to use Apache Spark are not supported. @@ -64 +60 @@ For information about estimating your capacity requirements, see [Determine capa - * Currently, the feature is available in the following AWS Regions: + * Capacity reservations is available in the following AWS Regions: @@ -96,0 +93,4 @@ For information about estimating your capacity requirements, see [Determine capa + * [Control capacity usage](./capacity-management-control-capacity-usage.html) + + * [Automatically adjust capacity](./capacity-management-automatically-adjust-capacity.html) +