I was wondering if something that we do (as suggested by our partners) with HP-UX and Progress and Oracle databases is coherent and applicable to a SQL Server installation.
We have an HP EVA6400 storage to deploy LUNs to all our DBs, VMs and stuff. It is an easy to use storage system, that uses all disks in the disk group at the same time for every LUN created. Easier to create and present LUNs, but lacks the fine grain that other storages have -- like, choosing _exactly_ disks 1, 2, 3, 4 and 5 to create a RAID 5, for instance.
What we do in HP-UX is create 4 LUNs for each mount point and striping them with LVM, creating one volume, and using multipath I/O. It's been like this since the first SAN deployed here, in 2007, but I've never made any benchmarks myself to see if it the performance indeed improves.
But I've never seen this in the Windows world, not even found any documents stating if it is a good thing or not.
Anyone has any thoughts?