2 Comments

Purely Analytical Purpose which doesn't require any change in the data - Yes. But as soon as it changes to transactional side, a hybrid approach works best, like parquet. It's a columnar storage but uses row groups to even divide the rows along with columns.

Expand full comment

Yes this is completely true, even for pure analytical purposes, where you need to update the data. I kept it out of the post to not confuse people. But I might add it in.

Expand full comment