Running a few VMs in VMware Server 2.0.2 on Ubuntu Linux 10.04. All works good most of the time. Except when one of them, for whatever reason, needs to do a lot of disk I/O. Then, all of the VMs grind to a halt until the disk activity is done.
Why is this? I don’t know. Obviously, if all of the VMs are using the same disk, there’s going to be some contention for disk access. But, it seems that this slowdown is way worse than it should be.
Anyway, there is a solution to greatly reduce the impact of the problem.

Set all of your VMs to use only one CPU! Yeah, that’s right. I don’t know why that makes the situation any better. Of course, there will be less CPU resources available for your VMs, but it’s worth it if they don’t stop responding whenever one of them decides that it needs to use the disk for more than a few seconds.
I got the idea from a post on the VMware Server forums and sure enough it worked fine for me. Now the VMs are much more responsive while the disk is under load. I don’t know if this problem also plagues VMware Server running on Windows, but chances are you can fix it the same way if it does.