• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

ESXi passthrough with incompatible device?

Well, this sucks. Did a crystal disk mark from my win7 box over gig-e to the OI VM. Read is about 88MB/sec (good), but write was only about 32MB/sec. Natively, this hits about 88 too. So, still about a 50% hit, even with raw mapped disk luns.
 
@danswartz: just curious. Are your "physical" or "native" numbers based on having the server running as a VM with the HBA on directpath under ESXi? Or is it with the OS running bare-metal on the hardware (no hypervisor)?
 
Sorry for any confusion. Terminology I used: "VM" == running as a VM under esxi with raw mapped luns (not directpath, since my mobo/cpu do not support that). "physical/native" == running on the bare metal as installed there by the OI cdrom (or zfsguru cdrom). Does that make more sense?
 
danswartz said:
Sorry for any confusion. Terminology I used: "VM" == running as a VM under esxi with raw mapped luns (not directpath, since my mobo/cpu do not support that). "physical/native" == running on the bare metal as installed there by the OI cdrom (or zfsguru cdrom). Does that make more sense?
Click to expand...

Yes, it makes more sense. But it does beg a new question: there are more variables in play between your scenarios than just the disk IO between these two senarios. There are artifacts of the ESX cpu scheduler, networking and other overheads that might be in play in the comparisons.

Would it be too much trouble to compare OI as a VM using ESX disk simulation vs OI as a VM using directpath passthrough? That way all else is equal and you are just comparing the disk IO performance.

(of course, i can tell you that I have done this in the recent past and your 50% delta is about right...but it would at least take these questions off the table).
 
I cannot test with directpath passthrough - don't have either mobo or cpu that support it. I assume you mean vmdirectpath, which is the vt-d pci passthrough stuff? I have tried vmkds on vmfs3 one per physical sata disk and the raw mapped luns - no real performance difference.
 
Last edited:
You must log in or register to reply here.
Back
Top