Windows – VMware VM Generates 100% Disk Activity

What exactly it means you probably have to ask the developers or look at the code. But the concept itself seems to be similar to what Resource Monitor displays as Queue length in its graphs:

which manifests itself as a percentage in the main view as well:

Basically this represents how many hard disk commands are waiting to be fulfilled. If there is little I/O, then the disk can read/write fast enough to satisfy them all quickly, but if many things are to be done at various places on the hard disk there will be lots of waiting for the arm to reposition, etc. (Side note: above screenshots are from an SSD which suffers from the problem a little later).

Once the queue length rises above 1 you'll usually start feeling the system slow down. Especially so if it's the disk where your pagefile is on. Since this is a very low-level thing it affects everything and even the commands to read a faulted page from the pagefile will have to wait. If you thought performance went down when paging starts due to full RAM ... this is a lot worse (and it happened to me every time I made an SVN update in a large repository).

When using virtual hard drives, regardless of use with or without a virtual machine, you can improve performance by hosting the virtual hard drive file on it's own physical drive to avoid IO being lost, because the host OS is using it.

You can also improve performance by using a pre-allocated space, instead of a dynamically expanding virtual hard drive.

When using virtual drives it's important to remember that file fragmentation can occur twice. Once on the host, to the virtual hard drive file, and again on the files in the virtual hard drive; defragmenting both can improve performance if there is lots of fragmentation.

Page files, and swap partitions will heavily slow a virtual machine unless they are stored on a very fast drive. Generally it's better to give the VM as much ram as possible, and then do not use a page file/swap partition.