----- "Gary Bowers" <gbow...@itrus.com> wrote: > My experience with direct connected iSCSI storage on a TSM server is > that it gets abysmal performance unless you turn off Direct IO in > TSM. See other posts for that. It is technically possible, but with > the iSCSI limitation you might not want to use RMD "Raw Device > Mapping" in VMware. I am not sure on this, but it makes sense given > what I have seen and read about here. By the way, NFS and CIFS were > equally bad performers for disk pools with DirectIO turned on. They > seem to really need the filesystem caching. I'm "guessing" that > putting the disks in a VMFS would help buffer the writes, and give > you > decent performance. > > It is something that would need to be tested first. I'm confident > that it would be much faster than WAN connection back to the States. > Yuck. > > Good luck, > > Gary Bowers > Itrus Technologies > > On Mar 11, 2010, at 1:18 PM, Ochs, Duane wrote: > > > Good day everyone, > > Has anyone explored using TSM server (windows) on a VM using Iscsi > > storage ? No library requirement at this time. > > I have multiple European sites within close proximity of each other > > and they have outgrown the WAN coming back to the states. > > Only storage available there is Iscsi and they have a substantial > > VMware implementation which would allow us to ride on a VM if > > feasible/functional. > > > > Thoughts ? > > > > Thanks, > > Duane
You probably would want the iSCSI storage linked back via VMWare and .vmdk image files so that Windows has no idea about it being iSCSI or otherwise - just a disk. Gives you more flexibility in the long run, but you would want to test both Direct IO and not to see which performs best in your configuration. I would have thought leaving DIO on would be the best if there's another OS doing filesystem cache somewhere else, but could be wrong there ;)