Virtual Warehouses

Updated Jan 14, 2022 ·

Overview

If multiple teams use the same warehouse, queries may start to queue and slow down.

The solution is to create separate warehouses for different teams or workloads. This way, there is no contention and each workload can run independently without blocking others.

Warehouse Types

Snowflake provides different warehouse types depending on workload requirements.

Warehouse Type	Description	Common Use Cases
Standard	General-purpose warehouse for most workloads	SQL queries, BI dashboards, and data loading
Snowpark-Optimized	Higher memory and compute configuration	Machine learning and memory-intensive processing

Standard warehouses work for most environments while Snowpark-optimized warehouses are mainly used for workloads that require additional memory and compute flexibility.

Warehouse Sizes

Warehouses come in multiple sizes depending on workload demand.

Size	Typical Usage
X-Small	Development and light queries
Small	General analytics and testing
Medium	Moderate production workloads
Large	Heavy analytics and reporting
XL and Above	Very large or complex processing

Larger warehouses are not always more expensive overall. A larger warehouse may complete the same query faster, resulting in similar total credit usage.

• Small warehouses run longer
• Larger warehouses complete faster
• Cost may stay similar depending on runtime

The goal is to find the smallest warehouse that delivers acceptable performance without wasting resources.

This is commonly called the "sweet spot" for warehouse sizing.

Auto-Suspend and Auto-Resume

Snowflake warehouses can automatically stop and start to reduce costs. There are two key settings:

• AUTO_SUSPEND stops idle warehouses
• AUTO_RESUME starts warehouses automatically

In the example below, the warehouse automatically suspends after 5 minutes of inactivity.

CREATE WAREHOUSE analyst_wh
AUTO_SUSPEND = 300
AUTO_RESUME = TRUE;

These settings are important because warehouses continue consuming credits while running, even if no queries are active.

Multi-Cluster Warehouses

Multi-cluster warehouses help handle high concurrency workloads. When demand spikes, additional clusters automatically start. When demand drops, they stop.

You can set the minimum and maximum number of clusters, and Snowflake manages the rest. This is especially useful when many users or applications run queries simultaneously.

CREATE WAREHOUSE engineering
  WAREHOUSE_SIZE = 'LARGE'
  MIN_CLUSTER_COUNT = 1
  MAX_CLUSTER_COUNT = 3
  SCALING_POLICY = 'STANDARD';

NOTE: Multi-cluster warehouses are available in Enterprise Edition and above.

Scaling Warehouses

Snowflake supports two ways to improve performance.

Scaling Method	Purpose
Scaling Up	Increase warehouse size for heavy queries
Scaling Out	Add clusters for many concurrent queries

Scaling up helps a single large query run faster. Scaling out helps many users run workloads at the same time without queueing.

Once multi-cluster warehouses are set up, you can choose between two scaling policies:

Policy	Behavior	Best For
Standard	Adds clusters immediately when queries queue	Unpredictable workloads
Economy	Adds clusters only if demand stays high	Stable and predictable workloads