In Snowflake, what is the difference between partitioning and clustering?

In Snowflake, partitioning and clustering are two distinct ways to organize data for performance. Partitioning is a physical data layout feature that splits a table into separate files on the storage layer based on a partition key. Each partition is stored in its own micro‑partition, which allows Snowflake to skip entire partitions during query execution if the filter matches the partition key. Partitioning is defined at table creation and cannot be changed later.

Clustering, on the other hand, is a logical re‑ordering of data within each micro‑partition. Snowflake automatically maintains clustering metadata and periodically reorganizes data to keep rows with similar values close together. Clustering improves range scans and reduces the amount of data scanned for queries that filter on non‑partitioned columns. It is optional and can be added or dropped after table creation.

In practice, use partitioning for high‑cardinality, frequently filtered columns (e.g., date or region) to eliminate whole partitions. Use clustering for columns that are often used in range predicates or joins but are not suitable for partitioning. Both techniques complement each other: partitioning reduces the number of micro‑partitions scanned, while clustering reduces the amount of data read within those partitions.