Partitioning Under the Hood in MySQL 5.5

We will primarily go through the implementation and how partitioning works inside MySQL server, rather than talk about how to use partitioning.

We will go through how partitioning works and explain how COLUMNS partitioning in 5.5 is different from the partitioning types in 5.1, and also talk about truncate partition and MyISAM key caches can now be used per partition.

We will explain why the current implementation have some challenges with tables with many partitions (too much overhead since it need to open and lock each and every partition before the optimization stage, where the pruning happens). And what is being done to solve it.

• Decrease the overhead of many partitions by using the new Metadata locking to allow pruning before locking (to also prune away open and lock operations).
• Exchange partition with table, swap place of a partition and a table.
• Explicit pruning, only use the partitions specified in the FROM clause
• Parallel ALTER TABLE where the partitions are altered in parallel, thus higher utilization of multi CPU systems.

At the end of the session we will also have questions and answers where the audience can get technical answers, and have the chance of getting their voices heard directly by the developers. Maybe by discussing new feature requests that we should focus on next (Global index, exchange partition with table, default list partition, explicit pruning, parallel operations such as alter/partition scan/index scan/insert/update/delete) or proposals from the audience.

So take the opportunity to make your voice heard, while getting deeper insight into the server!