Snowflake · ADA-C02
Virtual Warehouses
Virtual warehouses are the compute resources that execute queries, load data, and perform DML operations. Unlike traditional systems where compute and storage remain tightly coupled, Snowflake's warehouses operate independently from data storage, enabling organizations to provision compute precisely when needed and scale independently based on workload requirements. Each warehouse consists of multiple compute nodes working together as a cluster.
Virtual warehouses are the compute resources that execute queries, load data, and perform DML operations. Unlike traditional systems where compute and storage remain tightly coupled, Snowflake's warehouses operate independently from data storage, enabling organizations to provision compute precisely when needed and scale independently based on workload requirements. Each warehouse consists of multiple compute nodes working together as a cluster.
They exist to provide isolated, elastic compute resources that can be sized, started, stopped, suspended, and resumed based on workload patterns without impacting data availability or other concurrent operations. This isolation ensures different teams, applications, or workload types do not contend for resources.
Practical Implementation Patterns
- Dedicated warehouses for different business units to enable cost tracking and chargeback
- Tiered warehouse sizes for different query complexity levels
- Auto-suspend and auto-resume configurations to minimize costs during idle periods
- Multi-cluster warehouses that automatically add clusters during high-concurrency periods
- Resource monitors that prevent unexpected cost overruns from runaway queries