It is with great shame that I admit that the information in yesterday's post is not accurate. Yesterday, I spoke about the "Hyper-V Virtual IDE Controller (Emulated)" performance counters as though they would give you I/O statistics for virtual machines using the virtual IDE controller.
There are only two scenarios in which those counters will show anything useful. Either when the VM very first boots up, for about 1 second before the OS is completely loaded... or, if the Hyper-V VM integration tools are not installed.
See, when the virtual machine has its integration tools installed, then even when the configuration screen about says "IDE Controller," it's no longer an emulated IDE controller. It's a synthetic virtual controller, just like you would get if you used the SCSI controller on the virtual machine.
If the VM did not have the integration tools installed, and was thus using the actual emulated IDE controller, that would mean that the I/O would need extra steps in processing, and it would travel through its vmwp.exe process on the Hyper-V host. However, if the enlightenments are installed on the VM, the I/O travels through the VMBUS. This makes the I/O faster, and it explains why when using synthetic devices as opposed to emulated ones means that you can no longer see I/O happening in vmwp.exe, (it should travel straight to the "System" process on the root partition,) nor will you see anything but zeros in the "Hyper-V Virtual IDE Controller (Emulated)" performance counters.
To reassure you that I'm not just making s*$@ up this time, this MSDN blog post says basically the same thing. And so does this even better one.
"The next thing to notice is the “Fast Path Filter”. This is a filter driver that gets installed on all disk objects in the virtual machine – whether they are IDE or SCSI. It allows us to pass directly to the VMBUS based path for everything except low level disk operations (like partitioning a disk)."
So there you have it. But unfortunately, this means those counters I mentioned are pretty much worthless now. But all is not lost! You can still use the "Hyper-V Virtual Storage Device" counters, however, the counters seem to be about how much I/O is being done on the VHD/X of the virtual machine, and not the precise I/O being done by the virtual machine itself.
Not as good, in my opinion, but it's still something.
Also, I'd like to thank Chris S from Serverfault for enlightening me (pun fully intended) and putting me on the right path.