Re: Cleanup proposal: Is the "sql-bench" subtree still useful in binary packages?

[Michael Widenius]

Hi!

>>>>> "MARK" == MARK CALLAGHAN <mdcallag@stripped> writes:

>> It's true that the comparisons results are not very usable anymore,
>> but the sql-bench test is still a very good test to include:
>> 
>> - It's tests totally different things than mysql-test-run
>> - It can be used to very fast test for regression in base code.
>>  (We plan to make this part of our build system to test for
>>   regression and I suggest you do the same!)
>> - It's also very easy to use for end users as a base to do a fast test
>>  of the speed of some specific feature and how different options
>>  affects this.

MARK> I think sql-bench is underutilized. I will begin using it for my local
MARK> builds as a performance regression test. It might also be useful to
MARK> share results between developers. For dbas, it would be good to
MARK> improve it to a point where it could do basic performance testing to
MARK> compare different hardware and configurations.

We are just working to set up a system where one can trivially run
sql-bench with a lot of different configure and mysqld startup
options.

This will be published at:
https://launchpad.net/mariadb-tools

We are planning to do 20+ different runs on MariaDB the next 2 weeks and
publish the results. (Almost all the results are of course relevant
for MySQL too).

When we publish the results we will also publish more information
about the test runs and ask for suggestions for other things to test.

Regards,
Monty