Most people I know who have spent a lot of time with Hyper-V have had the experience of accidentally taking too much memory away from the parent partition.  This happens when they start too many virtual machines – and all of a sudden the performance and responsiveness of the parent partition goes down significantly.

The response from people who hit this is usually to stop the last virtual machine that they started, to reduce its memory, and then start it up again.

This solution has worked in the past – but is no longer an option with dynamic memory.  The reason why this will not work is because if you stop the last virtual machine – you just leave the memory available for your other virtual machines to use – getting you straight back where you started.

For this reason, with dynamic memory we have implemented a new “parent memory reserve” for Hyper-V.  Here we attempt to calculate an appropriate amount memory to keep for the parent partition and ensure that virtual machines with dynamic memory enabled cannot eat into this reserved memory.

That said – this reserve is not perfect.  Specifically – it only accounts for the memory requirements of Hyper-V in the parent partition.  If you are running other workloads in the parent partition (contrary to our best practice guidance) then you may still see parent partition memory starvation.

To help mitigate this issue – we are also providing a new registry entry that lets you override our parent memory reserve with your own static memory reserve.  This registry entry does not exist by default – but if you go to the HKLM:\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Virtualization registry key and create a new DWORD entry with a name of memoryreserve – you can then set the value to the static amount of memory that you want to reserve for the parent.

You should be warned – if you set this value too low; virtual machines will be able to use too much memory and cause performance issues for you.  Equally – the higher you set this the fewer virtual machines you can run.

That said – I have found this to be very useful on my laptop – where I use Hyper-V like a desktop operating system.  By setting this value to 2048 (on my laptop with 8GB of RAM) I can run multiple virtual machines and know that they will not cause Outlook / Internet Explorer in the parent partition to be affected.

Cheers,
Ben

Jeff has discussed this at some length over on the virtualization team blog, but as a general rule of thumb we believe that it is much better to have paging occur inside the guest operating system rather than at the virtualization layer (if paging is needed).

The simple reason for this is that the guest operating system has far better understanding of which are the best sections of memory to page out – where as all the virtualization layer can do is to guess at what should be paged out.

In my recent demonstrations of dynamic memory I came upon another interesting angle to consider.  Here is a screenshot of task manager from a virtual machine that is running at –23% memory availability (i.e. it does not have enough memory available and is paging in the guest):

What is fascinating about this screenshot is that even though this virtual machine is significantly short of memory, the guest operating system is still keeping 112mb memory available / as file cache.  The reason for this is that the copy of Windows inside the virtual machine knows that even though it is short of memory – it can provide the best experience for the user of the virtual machine by not using all the memory that it has and by keeping a little bit free to serve as a cache / be there for new applications.

It is exactly this sort of logic that gets lost when paging is done at the virtualization layer instead of inside the guest operating system.

Cheers,
Ben

Malware Unmasked 2 from OEM Sales on Vimeo.

With Windows Server 2008 R2 SP1 we have added a number of new performance counters for dynamic memory.  These counters allow you to get some extra insight into what is actually happening with memory on your Hyper-V server.  The counters are grouped in two categories.  First is “Hyper-V Dynamic Memory VM”:

These counters allow you to access specific information about dynamic memory inside each of the virtual machines on your system.  The second category is “Hyper-V Dynamic Memory Balancer”:

These counters allow you to view information about dynamic memory across the entire physical computer.

One thing to note is that unlike the user interface – which uses the concept of “memory availability” (which I discussed yesterday) – the performance counters instead talk about memory pressure.  This is measuring the same thing as memory availability (i.e. how much memory does the virtual machine have compared to how much memory does it need) but the math is slightly different.  A virtual machine with a pressure of 100 has exactly the amount of memory that it needs (this is equivalent to a memory availability of 0%).  If the virtual machine pressure goes over 100 – it now has less memory than it needs (equivalent to a negative memory availability) and if the virtual machine pressure is under 100 then the virtual machine has more memory than it needs.

Virtual machine pressure is simply calculated by taking the amount of memory the virtual machine wants, dividing it by the amount of memory the virtual machine has and then multiplying the result by 100.

My favorite performance counter is the “Average Pressure” counter under the “Hyper-V Dynamic Memory Balancer” category.  This gives you a very simple view of the overall memory allocation of your system:

As long as this number is under 100, you know that there is enough memory is your system to service your virtual machines.  Ideally this value should be at 80 or lower.  The closer this gets to 100, the closer you are to running out of memory.  Once this number goes over 100 then you can pretty much guarantee that you have virtual machines that are paging in the guest operating system.

Cheers,
Ben