Why Snowflake separates compute from storage - Visual Breakdown

Process Flow - Why Snowflake separates compute from storage

Data Stored in Central Storage

↓

Compute Clusters Access Storage

↓

Cluster A

↓

Process Queries

↓

Results Returned

Snowflake stores data centrally and lets multiple compute clusters access it independently to run queries.

Execution Sample

Snowflake

CREATE WAREHOUSE WH1;
CREATE WAREHOUSE WH2;
USE WAREHOUSE WH1;
SELECT * FROM table;
USE WAREHOUSE WH2;
SELECT * FROM table;

Two compute warehouses run queries on the same central data storage independently.

Process Table

Step	Action	Compute Cluster	Storage Access	Result
1	Create Warehouse WH1	WH1 created	No data accessed	Ready to run queries
2	Create Warehouse WH2	WH2 created	No data accessed	Ready to run queries
3	Run query on WH1	WH1 active	Reads data from central storage	Query result from WH1
4	Run query on WH2	WH2 active	Reads data from central storage	Query result from WH2
5	WH1 and WH2 run independently	Both active	Shared storage accessed concurrently	Results returned separately
6	Stop WH1	WH1 stopped	No storage access	No queries run
7	WH2 continues	WH2 active	Reads data from storage	Query results continue
8	End	WH1 stopped, WH2 active	Storage remains intact	System stable

💡 Execution stops as warehouses are stopped or queries complete; storage remains separate and persistent.

Status Tracker

Variable	Start	After Step 1	After Step 2	After Step 3	After Step 4	After Step 6	Final
WH1 State	Not created	Created	Created	Active	Active	Stopped	Stopped
WH2 State	Not created	Not created	Created	Created	Active	Active	Active
Storage State	Data stored	Data stored	Data stored	Accessed	Accessed	Accessed	Accessed

Key Moments - 3 Insights

Why can WH1 and WH2 run queries at the same time without interfering?

Because both warehouses access the same central storage independently, as shown in steps 3 and 4, they do not block each other.

What happens to data when a compute warehouse stops?

Data remains safe and unchanged in central storage, as seen in step 6 where WH1 stops but storage is still accessed by WH2.

Why is separating compute and storage beneficial?

It allows multiple compute clusters to scale and run queries independently without duplicating data, demonstrated by WH1 and WH2 running separately.

Visual Quiz - 3 Questions

Look at the execution table, what is the state of WH1 after step 6?

A Active

B Stopped

C Not created

D Paused

💡 Hint

Check the 'WH1 State' in variable_tracker after step 6.

At which step do both WH1 and WH2 run queries concurrently?

A Step 3

B Step 4

C Step 5

D Step 6

💡 Hint

Look for the step where both warehouses are active and accessing storage.

If storage was not separated, what would happen when WH1 stops?

A WH2 would also lose access to data

B WH2 would continue normally

C Data would be duplicated

D Queries would run faster

💡 Hint

Think about the benefit of separate storage shown in steps 6 and 7.

Concept Snapshot

Snowflake separates compute and storage.
Data is stored centrally and persistently.
Multiple compute warehouses access data independently.
Compute clusters can scale up/down without affecting storage.
This design improves concurrency and cost efficiency.

Full Transcript

Snowflake stores data in a central storage layer. Multiple compute warehouses can be created to run queries independently. Each warehouse accesses the same data without copying it. This separation allows warehouses to scale and run queries concurrently without blocking each other. When one warehouse stops, data remains safe and accessible by others. This design improves performance and cost by separating compute resources from storage.