Hi all,
we have a brand new two hosts Hyper-V cluster, Dell R610 (dual Intel Xeon E5645 2,40GHz 6 cores, 96 GB RAM) with MD3220i iSCSI storage (12 x 500 MB near line SAS, Windows 2008 R2 SP1, one Clustered Shared Volume about 4 TB. Two dedicated NICs in each host for
iSCSI, jumbo frames, dedicated switches, MPIO. TCP Offloading and Receive Side Scaling are disabled, preferred Network for Cluster Shared Volumes Communication has been configured, HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters\Interfaces\<Interface
GUID>\TcpAckFrequency set to 1 for NICs to iSCSI storage.
Disk performance in VMs is very poor, and overall VM performance is unacceptable (just as an indication, opening Server Manager takes 30 seconds or more). I have tested disk performance with HD Tune Pro and compared them with a very similar infrastructure on VMware and I get more than three times as much average access times, up to ten times max access times, one third or less IOPS and average speed.
I really don't know what else to look into. Customer is starting to think that moving to VMware would be a more appropriate setup.
I also have to mention that at first setup network performance on the hosts was very low and operations like quick formatting a disk on the iSCSI storage were taking minutes instead of seconds and producing time out errors on iscsiprt in Event Viewer. That has been resolved disabling C States in BIOS (http://support.microsoft.com/kb/2000977) and TCP Offloading capabilities and Receive Side Scaling in the host OS.
I would really appreciate any suggestions.
Thanks, Alessandro