Storage-contained
Now typically in a use case-oriented partitioning strategy you don’t want to select a column that has a large number of potential value types. So for example you wouldn’t want it to be a person's User ID. This would just create an explosion of little files and defeat the purpose of partitioning the data, right?
Wellllll yeah… there’s another way of partitioning that sort of turns this on its head. This second method is a storage-contained method of partitioning. In other words, what dictates the separation of one partition to the next is literally a cap on how much-allocated space each partition has, and that allocated space is tiny. I mean if the data were uncompressed it would be between 50-500 MB, and with modern compression it might be in the teens of megabytes.
Watch the full video & read the whitepaper: https://www.intricity.com/learningcen...