I decided to benchmark the code that builds out the partitioned and non-partitioned tables.
The results are very perplexing.
| table-type | insert | select |
|---|---|---|
| non-partitioned | 34% | 1.02% |
| code-partitioned | 26.98% | 96.87% |
| mysql-partitioned | 38.89% | 2.03% |
Code-partitions are fast on the inserts.
MySQL Partitions are fast on the reads.
This seems to suggest a complicated MySQL master-slave set up where you just have code partitions on the Master and MySQL partitions on the slave. It’s probably not worth the effort setting up, so I’ll just go with the MySQL partitions and work on finding ways to make the MySQL code work faster.
What do you folks think?
One reply on “More MySQL 5.1 Benchmarks: My Code is Faster? And Slower? WTF?”
[…] wrote benchmarking software about 2 years ago to test partitions but I’ve since abstracted the code to be database […]