Athena provisioned capacity model doesn't utilize the entire resource

Question

Hi there,

I ran a workload on Athena's capacity reservation setup which completed successfully but didn't utilize the entire resource in the reserved capacity. It only used 78% percent of it consistently throughout the entire run (a span of 35 minutes). On top of that, I repeated the experiment which yielded almost the same results (78% utilization for 35 minutes). I'm wondering why it capped at that number.

Provisioned capacity: 160 DPUs
Utilized capacity: 125 DPUs

Answer

Athena allocates resources dynamically based on overall service load and query volume. Queries go through a queuing process before being assigned resources.

The queuing system aims to distribute resources efficiently across all customers. It's possible your workload was not large enough to fully utilize the reserved capacity.

Factors like query complexity, data size and distribution of requests over time can impact how much of the reserved capacity gets utilized at a given time.

Utilization levels tend to average out over longer periods as workloads and queues fluctuate. Running the workload repeatedly may give a more accurate picture of long term utilization.

You can check Athena documentation or contact support for guidance on sizing capacity reservations based on expected query patterns and SLAs. Over-provisioning is minimized through Athena's flexible resource allocation.

The reservation ensures your important workload always has priority access to resources, even during peak usage. But full utilization at every moment is not guaranteed.

Athena provisioned capacity model doesn't utilize the entire resource

Relevant content