Comments on: Partitioning, Free Lunches, and Indexing https://www.percona.com/blog/partitioning-free-lunches-and-indexing/ Sat, 22 Jan 2011 18:05:08 +0000 hourly 1 https://wordpress.org/?v=6.5.2 By: Justin Swanhart https://www.percona.com/blog/partitioning-free-lunches-and-indexing/#comment-10902396 Sat, 22 Jan 2011 18:05:08 +0000 http://www.tokutek.com/?p=1747#comment-10902396 If you have a table with a long tail, which is rarely deleted from and old data is rarely accessed, then partitioning keeps the indexes on the hot partitions smaller which is results in better performance. This is quite common, and the situation in which I most often suggest partitioning.

You can also use innodb_file_per_table w/ symlinks (really only safe with the facebook patch) to place partitions on different IO subsystems. If you execute queries in parallel (see http://code.google.com/p/shard-query) then you can get much better performance using partitions than not using them. This is an example of ‘divide and conquer’.

]]>