SAN Performance

I have a situation where a client is experiencing performance issues with our application which heavily relies on SQL. We have a very large table that has heavy fragmentation, so to ensure this is rectified first before exploring other performance related issues we have decided to rebuild the index. We have had to allocate additional storage to attempt this by adding 3 LUNs (500gb each) + 3 additional SQL data files.When running the rebuild I monitored the I\O and noticed the throughput was very poor.The clustered index in question was sitting on 2 files and when rebuilding the index you can see the 3 new data files being populated (These files have been pre sized).On all three files I am only getting 1-3 I\O per second and under 1mb\sec throughput.When comparing this on a fixed disk system I can see I\O’s are far greater.When using SQLIO against each LUN on the SAN the performance is very respectable.Has anyone had any similar issues with performance and SANs.I did notice the driver for the HBA on this system was relatively old and a newer one is available, would anyone think upgrading this may address the issue?Any ideas would be much appreciated.This definitely appears to be a storage issue.

When you do SQLIO tests are you testing all LUN's at the same time and using a file size that makes the total size of test files larger than the cache available to you from the SAN?

WesleyB

Visit my SQL Server weblog @. http://dis4ea.blogspot.com

|||I have tested each LUN individually and yes the test files are larger than the cache.|||

I would try to test all LUN's at the same time since this might give a different result.

WesleyB

Visit my SQL Server weblog @. http://dis4ea.blogspot.com