I've read the articles on partition alignment and seeing the potential gains of 20-30% was quite keen to try them out.
We use a SQL2005 environment over WS2003, however, the original filesystems were installed over W2K and SQL2K so it looked a fair bet our partitions would be out of alignment.
My first step was to test each drive array to see which ones were in and which ones were out of alignment, this was pretty straight forward and as expected most were in need of attention :-)
Before doing anything we wanted to get a base line so we could measure the improvement and to be honest wave a little flag to the business ... after reading the articles several times along with several blogs it seemed that using a batch file to run various SQLIO tests was a decent way to go.
The information was collated, our test server was then aligned and the tests were run again - the difference was not showing as 20-30% gain
I did the same type of test on other arrays - RAID1 pairs, RAID1+0 sets and the results were inconsistent.
Is there a better way to get reliable performance results? e.g. results that relate to real world gains or losses as our tests don't back up the official Microsoft data so we're left in a quandary over whether it's worth implementing the changes to production systems.