I've recently loaded ESX4 Update2 on two whitebox servers that I'm using for data backup and DR purposes. Both have 64-bit Quad-Core Intel E5405 CPUs, plenty of RAM, supported SATA controllers, and on-board supported Intel NICs.
I did a P2V migration for the OS that originally was the base OS (Windows Server 2008 SP1) on Server A, and it is now a VM on that server. It was also running on Server B. That OS is running a network monitoring tool (PRTG) that I use to keep an eye on my network, including latency within my network, and to my remote sites.
What I've noticed is that, when the Windows VM is running on both Server A and Server B, I see surprisingly variable ping latency to both the service console on both servers, as well as from the VM out to the various hosts on the network. I thought initially that it was a WAN issue, until I realized that the latency was being experienced from the VM even within the LAN, and even from the VM to the IP address of the service console for its host ESX hypervisor.
Other pings between hosts on the LAN show <1ms latency with little variability, so the issue seems isolated to the ESX environment.
What could be causing this, and what can I do about it? I don't see evidence of poor networking performance either in the service console environment or the VM environment -- file transfers are speedy enough, both when moving the VM between servers last night and copying files via Windows networking.
PRTG supports having remote probes, so I could move the probe portion of my network monitoring to another non-VMWare server, but I'd like to understand this issue before I resort to that. I've seen some discussion related to network latency in older versions of VMWare, but nothing with 3.5 of 4.0.
If your source machine previously installed with antivirus, I recommend you to uninstall it first before you P2V the machine. I've had similar experienced before with machine installed with Symantec & and M'Cafee AV as I explained HERE.
BTW, the workaround for P2Ved machine, uninstalled VMware tools, antivirus(M'Cafee) or disable network firewall (Symantec). Then reinstall VMware-tools, monitor ping result then reinstall your AV.
Thanks. I wasn't even aware that there were different flavors of vNics. I'm learning as I go here.
I went with the vxmnet3 driver and it seems to have improved things significantly.
The VM is doing 100mb/sec backup at the moment (rate-limited by the domain controller's NIC) and I'm seeing an average of 1ms instead of 13ms (mostly because of a high stddev in the latency values in the previous E1000 vNic setup). So definitely a difference. Still not ideal, but I'll let it bake in for a few hours and see if the difference pans out in terms of the monitoring data.
I really appreciate the helpfulness of folks here.