Why Microsoft doesn’t like client hypervisors

VMware has a server hypervisor, and VMware has announced a client hypervisor. Citrix has a server hypervisor, and Citrix has announced a client hypervisor.

VMware has a server hypervisor, and VMware has announced a client hypervisor. Citrix has a server hypervisor, and Citrix has announced a client hypervisor. Microsoft has a server hypervisor, so why haven’t they announced a client hypervisor? Are they being smart, stupid, or clueless?

I guess this depends on whether (and why) you think client hypervisors will be important.

I’ve written in the past that I think the big reason client hypervisors will be important is not because people will run side-by-side operating systems, but instead because the hypervisor “scrubs” the hardware, thus allowing a single disk image to be used on multiple types of client devices. (You get the management ease of VDI while letting users leverage local computing resources and work offline.)

If you agree that’s the reason client hypervisors will be big, then it’s no surprise that Microsoft is ignoring this space since they already have a solution that runs directly on client hardware. (see ‘Windows’) Back in the old days, you couldn’t just “port” one installed copy of Windows from machine to machine. (In other words, if you popped the hard drive out of one laptop and stuck it in another, there was a pretty good chance that it wouldn’t boot.) That was because (1) older versions of Windows were’t too good at dealing with unexpected new hardware, and (2) different laptop models were actually kind of unique and different back then.

But think of that same disk-swapping experiment today. I’ll bet if you pulled a Vista hard drive out of a random laptop and stuck it in another, there’s a pretty good chance that thing would work. And that’s because (1) as much as we all hate Vista, the PE / hardware recognition part of it is actually quite good, and (2) nowadays all the laptops have the same chips in them anyway. From a driver standpoint, there’s not too much variety out there.

Given all that, you can understand how Microsoft would be saying “Client hypervisor? Pfft! We don’t need a stinking client hypervisor! We’ll use Windows 7. Remember how I wrote about the client hypervisor becoming the new HAL? Microsoft would disagree..Microsoft wants the HAL to be the new HAL. ;)

If you narrow your focus enough…

The other ironic thing is that building a client hypervisor is actually a lot harder than a server hypervisor because you have to support everything on the box. A client hypervisor needs to expose the battery’s power state to the VM and provide direct USB access and GPU support and all sorts of tricky stuff. So how are the client hypervisor vendors achieving this goal? Sure, partly by building cool new technology, but a big part is also due to them narrowing the field of devices they’ll support. (So far it seems that everything will require an Intel CPU with V-Pro.)

The irony is that by narrowing your landscape of potential clients so much, you marginalize the need for a client hypervisor in the first place!

So with these narrow HCLs, how much do you want to bet that a plain Windows 7 image (with maybe two sets of graphics drivers and two sets of network drivers) will work on all these anyway?

This almost makes sense, but kind of not

Of course the “catch” in all of this is that this is assuming that a bare-metal client hypervisor is only used to scrub the hardware. If it’s actually used to run multiple VMs, well, then I guess none of this matters?

What do you think Microsoft’s play is here? Are they secretly acting aloof until they suddenly release Hyper-V for clients? (Hey, it worked for Hamlet!) Are they actually scared about losing control, or will Windows forever be Windows? Or will Microsoft just roll their eyes and try to invalidate and trivialize client hypervisors until they buy one and put the rest of their partners out of business?

(By the way, for this whole article, I’m talking about bare-metal Type 1 client hypervisors. This has nothing to do with Microsoft Virtual PC and their purchase of Kidaro which they’ve released as MED-V. MED-V is an application compatibility solution that happens to run in a VM on the client, and not relevant here.)

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

I may be an odd duck, but I want my client hypervisor for my laptop.  Not for portability but to run multiple OS images.  And in addition to people like me, you have the client hosted VDI scenario.  Surely Microsoft sees this and would get their heads out of the sand at some point.  If not, we'll get it from one of their partners.

Hmm.  Wonder which one they'd prefer to see us use...


As far as I am concerned, the only useful thing a client hypervisor has to offer is disk virtualization. If we look at Ardence today, they have two fundamental capabilities - one is to strip windows of instance specific information (sid, computer name etc etc) and the other is streaming of the disk image.

With VDI people will figure out how to build monolithic images or images (os + apps + personality) which can be assembled on the fly.

The client hypervisor provides the other piece i.e. stream the disk or store the disk image as a file.

If I want to run multiple OS version or apps in compatibility mode with a particular OS, solutions like vmware/unity and/or kidaro make more sense.

Adding a hypervisor to the desktops also adds another layer which now needs to be managed!

Additionally, as pointed out in the article, a desktop hypervisor needs to deal with a lot more devices and their nuances, which will itself make it a big beast.


If you look at Microsoft's history they are typically a year or two behind VMware or Xen when it comes to putting a product out on the shelf .  Now I'm not saying that VMware or Xen is years ahead of Hyper-V, but what I am saying is that Microsoft seems to let others test the waters and then if everyone starts jumping in the lake, then and only then will they look at putting out their own version of the product.  It seems as thought they don't like to take risks, and instead becoming the leader (which they can and should) they would rather be the follower or the me too.  You might think I'm being harsh, but I think I'm just being honest.  Can anyone name anything revolutionary to come out of Microsoft in the last ten years...hardly.

@Brian, One question about your statement: "everything will require an Intel CPU with V-Pro"

The V-Pro features are built into the motherboard, and is something that you can not add-on to an existing motherboard that did have it in the first place.  That being said, if I buy a motherboard which has V-Pro, do think that whenever whoever release the client hypervisor that I will have the ability to flash the ROM on the motherboard to now take advantage of this new feature?  Or am I stuck on repurchasing all new systems with the new version of V-Pro (V-Pro 2 maybe?) in order to get this technology?


I want to run mutiple OS's side by side for sure. I want client personalities to do different things, offload mutimedia when the protocols fall down, and enable employee owned devices like laptops. In other words a lot more flexibility at the client. The management is a side benefit. vPro everywhere is pipe dream and just dumb, it's already proven not to be secure, and nither are Type 1 hypervisors if you know anything about hardware level hacking. So MS is smart, in that they know Type 1 is not needed right now, Type 2 is a much better play that does not require vPro. Looking forward to your Moka 5 interview as they really get why people want this feature.


I thnk it goes a bit deeper than that, Brian.  I believe some inside of Microsoft do in fact see the future of the desktop/laptop having some sort of multi-VM capability, and it probably will be a client hypervisor.  However, as the OS company, they are surely looking at ways to leverage multi-OS capabilities within Windows itself, rather than a separate entity.  Just as Hyper-V is provided with Windows Server, so will a Client Hypervisor be provided with a future Window (8? 9?).  As we know, doing a client hypervisor right is a lot more arduous than a server one because of the myriad input and output devices that comprise the client.  So, it might be a while before we see a real client hypervisor from Redmond.


I think, Microsoft should simply buy Citrix, because they are no longer partners in the Virtualization Market. They are competitors. When they buy Citrix, integrate everything in one good Management Solution, they have everything, what is missing...


Microsoft already has undertaken efforts to make Windows 7 more "virtualizable," so to speak.  It can handle VHD files natively, just like physical disks.  You can install Win7 directly into a VHD file and even boot from the VHD file just as if it were a physical disk.  Booting natively off a physical disk will become a thing of the past.  Maybe they'll build in functionality that will let you boot into these VHDs from a Type-2 VMM.  

Like you said, if portability between the different virtualization scenarios is the goal, the key is scrubbing the hardware dependency.  MS might be moving there, but Citrix already is there.  With Project Independence (type-1 client), XenServer (type-1 server) and Project KXen (type-2, from Xen.org), the same VHD could be the basis for local boots, VDI and hosted VMM use (think Macs).  Throw in Provisioning Services and you have a nice way to standardize these disks around a single file.


Does is seem unthinkable that once application virtualization matures, there will not be a need to run Windows on the client side or the server side for that matter? The application, by nature of its virtualization environment (run-time interpretor) and the underlying "hyperV" (or whatever)  firmware will become OS independent/agnostic. Thus you won't need Windows to run Win32 applications any longer. The virtualization  firmware will essentially create an engine in which virtualized applications will execute either on the local PC or remotely over a thin computing protocol of some sort. The precursors to this evolution can be seen in Vmware hosting Windows server, Java, and again, application virtualization technologies that are basically stripped down Windows emulation shells.