Handle Unpredictable or Ad Hoc Queries with the Query Acceleration Service

As a customer-centric organization, Snowflake is always innovating to help maximize economic value for our customers. This includes continuously evolving and improving the single elastic performance engine that powers the Snowflake Data Cloud, helping existing workloads run faster, more efficiently, and with more transparency.

Various settings such as warehouse size, the maximum number of clusters, and now, the Query Acceleration Service, allow our customers to optimize their queries to be more performant and cost-effective. The Query Acceleration Service (QAS) provides an even more flexible experience for customers with mixed workloads or ad hoc queries.

By offering a burst of additional compute resources to provide performance for eligible queries, QAS gives customers the flexibility and elasticity to handle unpredictable workloads, at the customer’s discretion. It runs alongside the customer’s existing warehouse to help accelerate the scan and aggregation portions of queries, in near real time. Upon completion of eligible query operations or fragments, the additional resources are then relinquished so customers only pay for what they need.

In this case, the majority of the workload has low utilization with the occasional unpredictable spikes from individually large queries, which can result in cannibalizing system resources and preventing new queries from running. As a result, this can lead to increased queuing. A customer today can accommodate for these spikes by resizing their warehouse, but this requires close monitoring and management of the warehouse. Instead, a customer can maintain a larger warehouse footprint to accommodate for these spikes.

In order to provide a more effective solution that balances customer needs around performance and cost, QAS allows our customers to accommodate for these performance spikes and pay for what they need, when they need it. As a result, there is the potential to achieve a couple of downstream benefits:

• Reduced queuing and better concurrency due to improved query performance
• Reduced overall workload time

In August, we introduced the full public preview of this feature. We looked at overall workload performance across customers who enabled QAS. The average workload times for the queries that were accelerated compared to queries run before QAS enablement were reduced by an average of 32%, with virtually no cost difference in running their workloads. We have seen improvements across various workloads. Here are a few specific examples:

Today, we are excited to announce the general availability of QAS. This feature is available to all enterprise customers and above. However, we have made eligibility views available to all customers to help you determine whether your warehouses/workloads are eligible for QAS.
Visit our documentation to learn more about QAS.