Performance for Snowflakes
Performance in Snowflake is about running queries quickly at the lowest credit cost. Tuning involves reading the query profile, right-sizing warehouses, using clustering keys on very large tables, avoiding spilling to disk and applying search optimisation where it pays off. Resource monitors help keep spending under control.
Stage: Performance and Governance · Part of the Snowflake roadmap · Last reviewed 19 November 2026
What to learn in Performance
- Read a query profile to find the slowest operator and data scanned
- Spot remote disk spilling and fix it with a larger warehouse or better SQL
- Decide when a clustering key helps a large table
- Prune micro-partitions by filtering on well-ordered columns
- Set a resource monitor to alert or suspend at a credit limit
Performance interview questions
When should you scale a warehouse up versus out?
Scale up, meaning a larger size, helps a single heavy query that is slow or spills to disk. Scale out, meaning a multi-cluster warehouse, helps when many concurrent queries queue. Increasing size does not fix concurrency queueing.
What is partition pruning?
Snowflake keeps min and max metadata for each micro-partition, so a filter on a column lets it skip partitions that cannot match. Good pruning means fewer bytes scanned. Clustering data on commonly filtered columns improves it for large tables.
Where Performance fits in the Snowflake roadmap
Performance belongs to the Performance and Governance stage (advanced level). Run it well.
Other units in this stage
Back to the full Snowflake roadmap