Concepts for Snowflakes
Snowflake concepts describe its architecture and core objects. Storage, compute and cloud services are separate layers, with data held in compressed micro-partitions and queries run by virtual warehouses of different sizes. Databases, schemas, tables and warehouses are the objects you create first, and understanding the layers explains Snowflake's cost and scaling behaviour.
Stage: Architecture · Part of the Snowflake roadmap · Last reviewed 19 November 2026
What to learn in Concepts
- Describe the storage, compute and cloud services layers and what each does
- Create databases, schemas and tables and organise them sensibly
- Explain micro-partitions and why Snowflake needs no manual partitioning
- Size a virtual warehouse and configure auto-suspend and auto-resume
- Explain how multi-cluster warehouses handle many concurrent users
Concepts interview questions
Why does Snowflake separate storage from compute?
Data lives once in shared storage, while any number of warehouses can read it independently. Teams can scale or pause compute for each workload without copying data, and they pay for storage and compute separately.
What happens when a virtual warehouse is suspended?
It stops consuming credits and its local cache is dropped, but data in storage is untouched. With auto-resume enabled it starts again on the next query, so a short startup delay and a cold cache are the cost of the saving.
Where Concepts fits in the Snowflake roadmap
Concepts belongs to the Architecture stage (beginner level), which takes about 20 hours. How Snowflake works.
Other units in this stage
Back to the full Snowflake roadmap