Performance regression in SQLbench
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Drizzle |
Fix Released
|
High
|
Brian Aker | ||
7.0 |
Fix Released
|
High
|
Brian Aker |
Bug Description
Starting back around build 1868 (October 21st 2010) we started seeing several tests in SQLbench have significantly higher run times (less is better in SQLbench, we are timing how long a test takes). I don't see anything in build 1868 or around that build which could cause this regression.
I focused on one test - update_
Will attach profile information as well as some debug information
This log file shows the results of update_ of_primary_ key_many_ keys and the time it is taking for each statement. Note that it takes 15-30 seconds to execute each statement at the beginning of the test but then around the 128th interation (of 256) the time drops back to less than a second as you would expect for a normal run.