Why virtual warehouses control compute independently in Snowflake - Visual Breakdown

Process Flow - Why virtual warehouses control compute independently

User Query Arrives

↓

Virtual Warehouse Receives Query

↓

Warehouse Allocates Compute Resources

↓

Query Executes Using Allocated Compute

↓

Results Returned to User

↓

Other Warehouse Receives Query

↓

Allocates Its Own Compute Independently

↓

Executes Query Separately

↓

Returns Results Independently

Each virtual warehouse in Snowflake manages its own compute resources independently, allowing multiple queries to run in parallel without affecting each other.

Execution Sample

Snowflake

CREATE WAREHOUSE WH1 WITH WAREHOUSE_SIZE = 'SMALL';
CREATE WAREHOUSE WH2 WITH WAREHOUSE_SIZE = 'MEDIUM';
-- Run query on WH1
USE WAREHOUSE WH1;
SELECT COUNT(*) FROM SALES;
-- Run query on WH2
USE WAREHOUSE WH2;
SELECT COUNT(*) FROM CUSTOMERS;

Two virtual warehouses run queries independently, each using its own compute resources.

Process Table

Step	Warehouse	Action	Compute Allocated	Query Status	Result
1	WH1	Start Query	Small Compute Cluster	Running	N/A
2	WH2	Start Query	Medium Compute Cluster	Running	N/A
3	WH1	Query Processing	Small Compute Cluster	Running	N/A
4	WH2	Query Processing	Medium Compute Cluster	Running	N/A
5	WH1	Query Complete	Small Compute Cluster	Completed	Count from SALES
6	WH2	Query Complete	Medium Compute Cluster	Completed	Count from CUSTOMERS

💡 Both warehouses complete their queries independently using their allocated compute resources.

Status Tracker

Variable	Start	After Step 2	After Step 4	Final
WH1 Compute	None	Small Cluster Allocated	Small Cluster Active	Small Cluster Released
WH2 Compute	None	Medium Cluster Allocated	Medium Cluster Active	Medium Cluster Released
WH1 Query Status	Idle	Running	Running	Completed
WH2 Query Status	Idle	Running	Running	Completed

Key Moments - 2 Insights

Why does WH2 start allocating compute only after WH1 has started its query?

Each warehouse allocates compute independently when it receives a query, so WH2 starts allocation when its query begins, as shown in steps 2 and 4 of the execution table.

Can WH1 and WH2 share compute resources during query execution?

No, each virtual warehouse controls its own compute independently, so they do not share resources, preventing one query from slowing down another (see compute allocation columns in execution table).

Visual Quiz - 3 Questions

Look at the execution table, what compute size does WH1 use during its query?

A Large Compute Cluster

B Medium Compute Cluster

C Small Compute Cluster

D No Compute Allocated

💡 Hint

Check the 'Compute Allocated' column for WH1 in steps 1 and 3.

At which step does WH2 start running its query?

A Step 2

B Step 3

C Step 1

D Step 4

💡 Hint

Look at the 'Query Status' column for WH2 in the execution table.

If WH1 was resized to 'LARGE', how would the execution table change?

A WH2's compute allocation would change to Large Compute Cluster

B WH1's compute allocation would show Large Compute Cluster

C Both warehouses would share Large Compute Cluster

D No change in compute allocation

💡 Hint

Compute allocation depends on warehouse size per warehouse, see 'Compute Allocated' column.

Concept Snapshot

Virtual warehouses in Snowflake manage compute independently.
Each warehouse allocates its own compute resources when running queries.
This allows multiple queries to run in parallel without interference.
Warehouse size determines compute power allocated.
Compute resources are released after query completion.

Full Transcript

In Snowflake, virtual warehouses control compute independently. When a query is sent to a warehouse, it allocates compute resources based on its size. Multiple warehouses can run queries at the same time without sharing compute. This means one warehouse's workload does not slow down another. After queries finish, compute resources are released. This design allows flexible, scalable, and isolated query processing.