Snowflake's Adaptive Compute (currently in Public Preview) introduces a new warehouse type that eliminates manual sizing decisions. Instead of choosing warehouse sizes, configuring multi-cluster settings, or enabling Query Acceleration Service, users set just two parameters: MAX_QUERY_PERFORMANCE_LEVEL (a per-query compute ceiling) and QUERY_THROUGHPUT_MULTIPLIER (concurrency capacity). The system dynamically allocates resources per query from a dedicated account-level compute pool. Billing shifts to a query-based model with no idle charges. Existing warehouses can be converted in-place with zero downtime, and bulk migration tooling is provided. Currently limited to Enterprise Edition on select AWS regions.

5m read timeFrom medium.com
Post cover image
Table of contents
Snowflake’s newest compute innovation removes infrastructure decisions so you can focus on data, not knobs.The Problem with Traditional WarehousesWhat is Adaptive Compute?The Two Knobs You Actually NeedGetting StartedGet Kapil Arora’s stories in your inbox
203 Impressions