Pricing, UI & Virtual Warehouses
Snowflake · Warehouses & Performance
Theory
Snowflake has two main bills: storage, charged on the average compressed data stored per month, and compute, charged in credits while virtual warehouses run. A virtual warehouse is a named cluster of compute that executes queries and DML. It comes in T-shirt sizes from X-Small upwards, and each step up doubles both the compute resources and the credits per hour.
Warehouses bill per second while running, with a 60-second minimum each time they start or resume. AUTO_SUSPEND stops a warehouse after a period of inactivity and AUTO_RESUME starts it again when a query arrives. In Snowsight you create and monitor warehouses under Compute → Warehouses, and pick one for each workspace file from the context selector.
Why we use it
Compute is usually the largest part of a Snowflake bill. Understanding sizes, per-second billing and auto-suspend lets you get the speed you need without paying for idle clusters.
When to use it
Create separate warehouses per workload, for example one for ETL and one for BI, so you can size and monitor them independently. Start small, measure with Query Profile, and only size up when queries are actually compute-bound.
What a Virtual Warehouse Is
A virtual warehouse is a cluster of compute resources that executes your queries; it has no data of its own — it simply reads from shared storage, which is billed separately purely on data volume.
T-Shirt Sizing and Cost
Warehouses come in T-shirt sizes from X-Small upward, each size doubling the compute (and credits per hour) of the one below it. You're billed per second only while a warehouse runs, and it can auto-suspend when idle. New standard warehouses default to Generation 2 (Gen2) where it is available; Gen2 has its own credit rates and does not offer the 5X-Large and 6X-Large sizes.
Picking the Right Size
Start small (X-Small or Small) for exploratory queries and interactive dashboards, and reserve larger sizes for heavy batch loads or complex joins over huge tables — resizing is instant and doesn't require restarting a session.
How it works
- A query needs compute
- Is the warehouse running? Yes: Run the query on it. No: AUTO_RESUME starts it (60-second minimum billed).
- Credits accrue per second while it runs
- Warehouse becomes idle
- Idle longer than AUTO_SUSPEND? Yes: Warehouse suspends and billing stops. No: Stay running and keep billing.
- Next query resumes it again
Example
CREATE WAREHOUSE analytics_wh
WAREHOUSE_SIZE = 'MEDIUM'
AUTO_SUSPEND = 300
AUTO_RESUME = TRUE
MIN_CLUSTER_COUNT = 1
MAX_CLUSTER_COUNT = 3
SCALING_POLICY = 'STANDARD'
COMMENT = 'Warehouse for BI team analytics';
CREATE WAREHOUSE etl_wh
WAREHOUSE_SIZE = 'LARGE'
AUTO_SUSPEND = 60
AUTO_RESUME = TRUE;
ALTER WAREHOUSE analytics_wh SET WAREHOUSE_SIZE = 'LARGE';
ALTER WAREHOUSE analytics_wh SUSPEND;
ALTER WAREHOUSE analytics_wh RESUME;
SHOW WAREHOUSES;
SELECT WAREHOUSE_NAME, SUM(CREDITS_USED) AS total_credits
FROM TABLE(INFORMATION_SCHEMA.WAREHOUSE_METERING_HISTORY(
DATE_RANGE_START => DATEADD(DAY, -7, CURRENT_DATE())))
GROUP BY WAREHOUSE_NAME ORDER BY total_credits DESC;Key points
- Storage is billed separately from compute, on compressed volume
- Each larger size doubles compute and credits per hour
- Per-second billing with a 60-second minimum on each start or resume
- AUTO_SUSPEND pauses idle warehouses; a very low value can lose the local cache
- Warehouse size can be changed at any time without downtime
- New standard warehouses default to Gen2 where available; check its credit rates