Estimate required capacity Signs that more capacity is required Check for idle capacity

Determine capacity requirements

Before you create a capacity reservation, you can estimate the capacity required so that you can assign it the correct number of DPUs. And, after a reservation is in use, you might want to check the reservation for insufficient or excess capacity. This topic describes techniques that you can use to make these estimates and also describes some AWS tools for assessing usage and cost.

Estimate required capacity

When estimating capacity requirements, it is useful to consider two perspectives: how much capacity a particular query might require, and how much capacity you might need in general.

Estimate per-query capacity requirements

To determine the number of DPUs that a query might would require, you can use the following guidelines:

DDL queries consume 4 DPUs.
DML queries consume between 4 and 124 DPUs.

Athena determines the number of DPUs required by a DML query when the query is submitted. The number varies based on data size, storage format, query construction, and other factors. Generally, Athena tries to select the lowest, most efficient DPU number. If Athena determines that more computational power is required for the query to complete successfully, it increases the number of DPUs assigned to the query.

Estimate workload specific capacity requirements

To determine how much capacity you might require to run multiple queries at the same time, consider the general guidelines in the following table:

Concurrent queries	DPUs required
10	40 or more
20	96 or more
30 or more	240 or more

Note that the actual number of DPUs that you need depends on your goals and analysis patterns. For example, if you want queries to start immediately without queuing, determine your peak concurrent query demand, and then provision the number of DPUs accordingly.

You can provision fewer DPUs than your peak demand, but queuing may result when peak demand occurs. When queuing occurs, Athena holds your queries in a queue and runs them when capacity becomes available.

If your goal is to run queries within a fixed budget, you can use the AWS Pricing Calculator to determine the number of DPUs that fit your budget.

Lastly, remember that data size, storage format, and how a query is written influence the DPUs that a query requires. To increase query performance, you can compress or partition your data or convert it into columnar formats. For more information, see Optimize Athena performance.

Signs that more capacity is required

Insufficient capacity error messages and query queuing are two indications that your assigned capacity is inadequate.

If your queries fail with an insufficient capacity error message, your capacity reservation's DPU count is too low for your query. For example, if you have a reservation with 24 DPUs and run a query that requires more than 24 DPUs, the query will fail. To monitor for this query error, you can use Athena's EventBridge events. Try adding more DPUs and re-running your query.

If many queries are queued, it means your capacity is fully utilized by other queries. To reduce the queuing, do one of the following:

Add DPUs to your reservation to increase query concurrency.
Remove workgroups from your reservation to free up capacity for other queries.

To check for excessive query queuing, use the Athena query queue time CloudWatch metric for the workgroups in your capacity reservation. If the value is above your preferred threshold, you can add DPUs to the capacity reservation.

Check for idle capacity

To check for idle capacity, you can either decrease the number of DPUs in the reservation or increase its workload, and then observe the results.

To check for idle capacity

Do one of the following:
- Reduce the number of DPUs in your reservation (reduce the resources available)
- Add workgroups to your reservation (increase the workload)
Use CloudWatch to measure the query queue time.
If the queue time increases beyond a desirable level, do one of the following
- Remove workgroups
- Add DPUs to your capacity reservation
After each change, check the performance and query queue time.
Continue to adjust the workload and/or DPU count to attain the desired balance.

If you do not want to maintain capacity outside a preferred time period, you can cancel the reservation and create another reservation later. However, even if you recently cancelled capacity from another reservation, requests for new capacity are not guaranteed, and new reservations take time to create.

You can use the following services and features in AWS to measure your Athena usage and costs.

CloudWatch metrics

You can configure Athena to publish query-related metrics to Amazon CloudWatch at the workgroup level. After you enable metrics for the workgroup, the metrics for the workgroup's queries are displayed in the Athena console on the workgroup's details page.

For information about the Athena metrics published to CloudWatch and their dimensions, see Monitor Athena query metrics with CloudWatch.

CloudWatch usage metrics

You can use CloudWatch usage metrics to provide visibility into your how your account uses resources by displaying your current service usage on CloudWatch graphs and dashboards. For Athena, usage availability metrics correspond to AWS service quotas for Athena. You can configure alarms that alert you when your usage approaches a service quota.

For more information, see Monitor Athena usage metrics with CloudWatch.

Amazon EventBridge events

You can use Amazon Athena with Amazon EventBridge to receive real-time notifications regarding the state of your queries. When a query you have submitted changes states, Athena publishes an event to EventBridge that contains information about the query state transition. You can write simple rules for events that are of interest to you and take automated actions when an event matches a rule.

For more information, see the following resources.