Our product historically used an Oracle backend, and after implementing simple date based range partitioning we got 20-30 time performance increase in our queries. We had used MySQL in other more minor products we had developed, but we couldn’t move to it until partitioning was implemented. In Dec 05 we started to test MySQL partitioning, where partitioning syntax was supported but the optimisations were not in place. As soon as the optimisations we put in, we were able to run a side by side comparison against unpartitioned tables and Oracle partitions. Not only did we get in MySQL a 30+ times performance increase, but the MySQL version ran nearly twice as fast as the same Oracle configuration in a like for like test.

More...