How is quick bulk delete performed?

Off-Topic
1

I’d like to quetions about bulk delete.

When performing bulk deletions,
Oracle performs drop by partitioning the table as a date. Quick bulk delete is possible because no logs are left.

For memsql,
The table specifies the date as ‘Columstore Index Key’.
When I perform a deletion by setting a range of dates in the Where clause,

Is it fast because it is deleted as a unit by ‘sorted-Row-Segment-Group’?

We need to explain how fast it is by the mechanism.

2

Columnstore tables consist of compressed segments. Data is sorted across those segments that contain 1M rows each. If you sort a table on data (clustered columnstore key) and delete statement filter on date many segment either won’t have any matching rows or will only have rows that match the filter. This allows memsql either ignore or fully delete full segment without inspecting each row. That’s why deletes are very fast in this case.

3

If then…
I want to filter the date field in the column store table so that I can quickly perform a bulk delete operation.

Can I perform bulk deletions more quickly by configuring a larger segment size?
Is there an ideal maximum for the segment size?
(Set value above default(1024000))

Thanks.

4

What are you observing now? Are deletes in fact slow or it’s more of a theoretical question?

5

Yes, it is. I need to be able to explain this logically.
Oracle has a small storage unit of 8k 16k 32k, but the memsql is stored in 32mb of segments and compressed, so the bulk deletion is fast.
Is this correct?

thanks.