Columnar Db’s want stuff to be contiguous on disk, and deletes cause the rest of the data in that “block” to be rewritten (imagine deleting a chunk out of the m

Columnar Db’s want stuff to be contiguous on disk, and deletes cause the rest of the data in that “block” to be rewritten (imagine deleting a chunk out of the middle of an excel table: you’ve got to move everything else up). This in turn, creates read+write load. Modern OLAP db’s often support it, often via mitigating strategies to minimise the amount of extra work they incur: mark tainted rows, exclude them from queries, and clean up asynchronously; etc.