Hi SpringTown

Thanks for your reply, first a correction to the hardware used it seems to be 12 * 4 core, so 48 cores per server, and I'll get six one for each instance used today.

If I knew the IOPS it would not be that tricky. In that case one could just use SQLIO on old setup vs. new. But sadly I don't have a specific iops benchmark I need to satisfy, it's more soft thing like "the server should not be slower than when it was on physical hardware", compensating that the new hardware is 4 years newer.

The aproach for migratiing our existing cluster boxes on to the VM-ware platform is to move the VS of the cluster, into a newly installed instance of the same name as the old cluster VS.

For SAN performance I'm not that worried as the old HBA's were 2Gbit and the new ones is 8Gbit, thus quadrupling the potential throughput. The idea is to co host smaller low activity boxes with the SQL as memory and CPU load of the SQL permits.

I'm not a SAN wizard but as I can see the config we have used on the IBM V7000 is Raid5 but as I said with 150 spindles on fairly small disks 560 GB., But perhaps the LUN are stripped across smaller RAID5 arrays, I can't really figure out how the V7000 does this, as the LUN says striped and from the disk view they say Raid5.

For migration of existing boxes we have decided to keep all disk layout the same, to simply migration, i.e split disks (tempdb, data and log) before = the same after migration.

For the new layout I'm still think if not ones disk is the best, as the data would end up on the same spindles no matter how many LUN's I would create. And I can always later move tempdb anyway to a seperate LUN if I would see a performance issue.

Best regards

Søren Udsen Nielsen