How does auto clustering work in Snowflake?

Question

The auto clustering in snowflake seems very unpredictable. Last week, I have created a cluster key on three tables. All these tables are similar in nature, has similar ingestion pattern. The cluster key is also on the same two columns. What I noticed is for one of the tables the clustering is getting triggered more often than the other two tables. This particular table is the smallest in size from the other two tables. And again the ingestion pattern is same for all the three tables.

I wanted to know how does the clustering algorithm work. How does it know when it should trigger itself.

Thanks

Answer 1

These slides explain it pretty well, I think: https://www.slideshare.net/InfoQ/automatic-clustering-at-snowflake The goal is:

Reduce worst clustering depth [to be] below an acceptable threshold to get predictable query performance.