Re: Partitioning performance issues // 5.1

[Mattias Jonsson]

This is partly because of bug#37252, i.e. the overhead of open and locking all partitions before pruning.

For inserts the work is about the same for a partition as for a table, you can in theory get a small performance gain, since the indexes in the partition is smaller, the depth in an index is usually O(log(n)).

The primary gain of partitioning is to ease the management of data, i.e. group data in different partitions and then add and drop partitions quickly etc. And to use pruning to skip partitions that can not satisfy the WHERE clause to search through less data.

Partitioning also opens up the possibility to implement parallelism in queries of big tables.