What type of query will benefit from the query acceleration service?

The Query Acceleration Service (QAS) is designed to improve performance by offloading portions of a query's workload, specifically large table scans, to shared compute resources provided by Snowflake. It is most effective for "outlier" queries that consume significantly more resources than the typical workload. The ideal use case for QAS is a query that scans a large volume of data but has a highly selective filter (e.g., in a WHERE clause) that returns a small subset of rows. In this scenario, QAS can efficiently perform the scan and data pruning, reducing the amount of data that the primary virtual warehouse must process, thereby accelerating the query.

A. Queries without filters or aggregation: The Snowflake documentation explicitly states that queries with filters or aggregation are the ones that benefit most from the query acceleration service.
C. Queries where the GROUP BY has high cardinality: While QAS can help with the initial scan for an aggregation, its primary benefit is not in processing the high-cardinality GROUP BY operation itself, which remains a computationally intensive task for the virtual warehouse.
D. Queries of tables that have search optimization service enabled: The Search Optimization Service and Query Acceleration Service are distinct features that solve different performance problems. Search optimization is for fast point-lookups, while query acceleration is for large scans. The presence of one does not imply a benefit from the other.

1. Snowflake Documentation: Query Acceleration Service.

Section: "Identifying Queries that Might Benefit from the Service"

Content: "Workloads that might benefit from the query acceleration service have the following characteristics: Queries that scan a large amount of data and include filters that are very selective." This directly supports option B.

2. Snowflake Documentation: Query Acceleration Service.

Section: "Overview of the Query Acceleration Service"

Content: "For example, a query with a highly selective filter on a table that is not clustered might be a good candidate for the query acceleration service. The service can improve the performance of the scan by pruning the data that does not match the filter." This provides a concrete example aligning with option B.