Signs of this were this: if you started a network transaction sending from the GUEST, to another server, the (windows) GUEST would lock up every couple seconds, this locking up would affect other guests hosted on the same (linux) host, and even the host itself in some cases. The end result for us was scrapping this server, and rebuilding it, that fixed the locking up issues.

(we still had to install a nic driver, and disable tcp offloading to correct the other network issues)

This locking up behavior wasn't apparent when receiving a file transfer, only when sending, both operations were slow.

Sending a file from the Guest to the Host in this configuration was also TERRIBLY slow... (10kb/s)

The problem caused the Host, and other guests to exhibit similar behavior, it seemed to mess with the tcp/ip engine on the linux host.

The problem get's worse the more the network is utilized on the guests.

Shut down the guest operating systems, and the vmware services, install a new driver if possible (as I mention below) disable tcp offloading. Set the tcp segmentation offload to off:

(command line:)

$ sudo ethtool -k eth0
rx-checksumming: off
tx-checksumming: off
scatter-gather: off
tcp segmentation offload: off

$ sudo ethtool -K eth0 tso off

Then restart the host, Test file transfer to and FROM the host, with no guests started, if this is up to speed or better you've fixed your problem.

Start the guests up, test transfers to and from the guests, are they still transferring well? -if not you may need to rebuild these servers.

If the guests still exhibit poor behavior move them to a Windows Host environement with a single core processor - do they work now?

If yes the vm is fubared.

Why does the vm work in the windows environment - I don't completely understand it, but how windows accesses the tcp/ip engine is different than how linux does, so the problem doesn't happen on the windows host - although on the windows host you will still see excessive resource usage but the network transfer won't be as slow as on a linux host, or 'jittery'.

How to Test for a Network Problem or Corrupt VM: ................

During these tests start up taskmgr.exe (task manager) on the GUEST and watch the NETWORK tab... (jitter or stopping occasionally of the network transfer indicated you may have a bad VM)

You can also start a resource monitor on the HOST (linux box) and watch it's network usage.

These tests will tell you if you have a corrupt VM (windows guest), if you transfer a file from the guest to another server is it slow, how about from another server to the guest?

If you transfer from the Host, to another server is it slow, how about from another server to the Host?

If you transfer from the Guest to the Host over the network is it slow?

Are you connected through gigabit?

Is there a more specific Linux Driver for your network adapter?

If there is install the specific Driver, if there isn't disable TCP offloading like is talked about in the links above (or try doing both)

Also you may want to try re-installing an updated VMware tools onto the Guests in question, I have found this fixed some performance issues in the guest.

If you have any power saving features enabled on the processors such as step-down or quiet n' coo disable them...

If you are on a gigabit connection try a 100mbit connection - this may resolve your issues, if it does you have a driver issue.

Did you use any-any update to setup / patch your vmware install? This is known to cause this behavior, disabling TCP stack offloading will help this.

The High-load on the servers should go away once you resolve the network issues!

After you resolve the network issues reset the host, and it should iron itself out... and stop using so many resources.