Why client hypervisors will be a big deal. Hint: It’s NOT about running multiple VMs.

Client hypervisors are all the rage now. Both VMware and Citrix are working on them, as are startups Virtual Computer and Neocleus.

Client hypervisors are all the rage now. Both VMware and Citrix are working on them, as are startups Virtual Computer and Neocleus. However, I think there is one big misconception about these things…

Most people talk about the impact of client hypervisors being that they will allow client machines to easily run more than one VM at the same time. While this is technically true, I do not believe this is the most important reason that client hypervisors will exist.

To me, the BIG advantage of the client hypervisor is that it will hide the real client hardware from the guest OS, thus allowing a single Windows disk image to run on different devices from different manufacturers. In other words, the client hypervisor will become the new “HAL.” (That’s “HAL” as in “Hardware Abstraction Layer,” not that computer from the movies.)

The ability to run a single copy of Windows anywhere (or, more appropriately, the ability to boot a single Windows disk image file anywhere) is a big part of my VDI 2010 vision, and a client hypervisor is needed to make that work end-to-end.

And that is the "game changing” part of the vision.

That’s not to suggest that the ability to run multiple OSes on the client device won’t be cool or useful. But in terms of having the most impact on the industry, being able to boot a single “master” Windows disk image anywhere will have a much bigger impact than a client device’s ability to run more than one OS at a time.

So the next time you hear someone say, “Gee, I don’t get what the big deal is about client hypervisors. Who cares if I can run multiple VMs?”, you’ll know what to tell them:

  • Improvements in app virtualization will lead to a greater percentage of apps being virtualized, and thus able to be provided on-demand to personalize the environment of a user booting from a shared master disk image.
  • Improvements in user environment personalization and virtualization will lead to a greater percentage of user settings and data being separated and saved outside of the OS, thus enabling users to boot from a shared master disk image while still getting their own custom environment.
  • Both of these mean that if we can figure out how to provide a single, generic master Windows image to our users no matter where they are, their apps, data, and personality can be applied on-demand once they load the shared master image.

Server-based virtualization platforms and VDI are already helping us do this for applications and desktops that are executed remotely. Client hypervisors will help us do this for local OSes (either pre-streamed and cached or booted on-demand).

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

You definitively got the point! I think it is of tremendous importance for a future desktop environment to just manage a few numbers of different desktop images.


I agree with you and would like to add 2 importants things :

- whatever will the client hypervisor become, the true value is not on the hypervisor itself but more on the usage above (streaming, security control...). Citrix or VMware or XYZ hypervisor could all become real, stable, "fast and furious"... I hope they will have "open API" that will allow everybody to use them and celiver services on top of all of them! Taht's were the value will be... That's were I want to get the choice...

- I also truly beleive that Citrix client hypervisor will become the "most interessant one" just because he is based on the Xen.org open source community and this could definitly help when it come the time to had more hardware support (video, I/O, USB...)... If the hypervisor become the new HAL, you deport this problem down to the hypervisor...


I really agree with your point on a single image. One of the great stumbling blocks of corporate deployment of Windows Vista (and soon to be Windows 7) is maintaining a single hardware image. If we could remove this blocker through a client-side hypervisor, then we could dramatically speed up the migration process and significantly reduce the associated cost.


I can't wait to see it in action, this is when we might start to see real flexible computing resources - I posted my thoughts on how this might work early last year.

Future vision of VDI? What if the Hypervisor was cheaper and flexible enough to run on BladePC’s/BladeWS’s - could vMotion/XenMotion then switch on demand?


Right on, Brian.  One of the more challenging aspects of implementing a client hypervisor is virtualizing all of the user-visible functions on a PC that haven't traditionally been an issue on virtualized servers.  First among them are graphics and USB.  I won't use this forum to try to make a case for why Virtual Computer's approach is superior.  However, I would encourage folks to drill into the details on how I/O such as graphics, USB, disk, wired/wireless network, etc. are being virtualized (or not virtualized, as the case may be) by various client hypervisors to ensure that the approach truly achieves the single Windows image model Brian is describing.


Although I could be wrong, you can almost bet the first iteration of a client hypervisor will not be 'open', either in API or source code.  Instead of creating a industry group to develop a standard, Citrix and VMware have signed seperate deals with Intel.  They will first try to monitize the client hypervisor, as they did with current hypervisors, but it will eventually be a commodity.  And I'm sure Citrix will tout the open nature and wide adoption of Xen, but XenServer is closed source and barely has a foothold in the market.  Here again, two industry learders work against the best interests of their customers.


Rodd made the important point about present client hypervisor plans. You won't be running an image created for a VMware hypervisor on a Xen platform any time soon (and vise verse). The ideas great, just as with server hardware virtualisation, but these companies are out to make money with this stuff.

Correct me if I'm wrong, but if you want to set up a VDI solution allowing for offline use, you're going to be stuck with the same hypervisor on you client you use of your Server VDI infrastructure.. and we know what that is in most cases at present. That image you check out needs a similar environment to run in.


Correct me if I'm wrong...but wouldn't a client side hypervisor also give VMware the ability to stream an OS down to a client much like the way Citrix's Provisioning Server (Ardence) delivers an OS to a system?  Or is this already possible in VMware without the requirement of something like a Windows client running ACE?


Gavin, I am not sure if I would agree with your second statement. Citrix's XenDesktop supports an open architecture infrastructure design and I recall seeing high percentages of deployments of this VDI technology on ESX 3.5 rather than XenServer. If what you say is correct then Citrix would be digging themselves into a deep grave.

I do agree with with the fact that both vendors are out to make money, who isn't? But what I don't agree with is how. My take on this is that they will make their money off of their delivery and management solutions.

After all, Citrix's CTO Simon Crosby has been cited many times that the hypervisor is a commodity. I see no change here.

But this is of course all of my own opinion.


A few comments:

@David - XenServer can already run on PC class hardware, and supports XenMotion in the free version. If a Client Hypervisor was used, the image would simply be streamed to the bladePC on demand (Like it is now for XenDesktop,) but it would be the same image that is used for the VMs that have to share hardware, thus reducing  management effort.

@Rodd - I'm hoping that you are wrong. It would be great to see this technology being shipped on every new PC as a feature like v-Pro is (in fact the XenClient ties prety tightly into vPro so hopefully,,,)

@Craig - Not VMware because they don't (currently) have any streaming technology, but Citrix - That is a different story altogether as you know. No VMware can't do this today - and ACE isn't a hypervisor - it is a hosted virtualization platform.

My personal feeling is that even though VMware might be on the path to getting a client Hypervisor, that's only a small part of the solution - components like Citrix's PVS and other image management tools will prove to be key in adoption in the enterprise.


Hi Brian,

 Great topic and I am interested in seeing a lot of the back and forth banter about the openness of the first client hypervisors.  I won't write some self-serving advertisement about my company but my general thoughts are that the hypervisor itself will eventually become commoditized.  The ISVs and corporate developers (and IT desktop managers) should not have to worry about the underlying hypervisor.  In the perfect world, there will be building blocks that allow them to choose components of OSes and, by applying predetermined rules, have the computing environment build "on the fly" for the end user - creating the OS environment and any type of applications (either virtualized or physically installed).  Your assessment of the whole image creation & management process being a big resource drain on IT is very spot on and virtualization, both at the hardware and application level will greatly reduce this problem.  


Excellent points throughout, nice article!

One excellent use case where client-side hypervisor would shine is a locked-down Linux VM used as a sandbox web browsing platform.

Taking things a step further, the traffic for these web browsing VMs could be on an isolated VLAN that has no layer 3 connectivity to the corporate network. You could likewise have the "corporate" VM image (running in parallel on the client) with limited or no connectivity to the web.

This approach could really curb the headaches associated with viruses, root kits, etc.



I'm cool with the fact that XenServer can run on PC hardware today - but correct me if I've misunderstood - there is no facility at the moment in XenServer to leverage the Graphics GPU or USB? and in todays world these are almost mandatory for any project that is going to deliver a Desktop replacment to users?




This link beings up a good point even if it's not true as a product. Why the hell does anybody care about bare metal client hypervisors now. This is years away before machines can handle it. The ACE, MED-D, MOKA 5 guest hypervisor solutions offer a much broader set of machines one can actually use this on. you need a VT enabled machines to run this stuff, and why on earth is anybody going to buy an VPro machine? Apple support is also key, don't see Apple offering anything there for some time. I would argue that there is a step that can easily be exploited before we get to full bare metal which will take years, let's not let VMWare and Citrix fool us!


Sorry Brian I don't think you are seeing the big win here.

While I agree that there will be significant benefit from positioning the client hypervisor as HAL, this is hardly a 'game changing event', not from where I'm standing anyway.  Yes, it simplifies existing desktop management processes, but we've been doing managed desktops for what 20 years now, these are mature processes without any significant shortcomings in a well managed environment.  The client hypervisor is an improvement over the current situation but only an incremental one.

Where it is going to make a more significant difference is in the way we support mobile employees and integrate contractors onto the corporate network.  The BYOPC model currently championed by Citrix is a highly visible example of how the client hypervisor will change things.  Instead of the current model where we have to contend with streaming apps or suffer the performance deficits of RDP (or whatever) connections running back to centralised environments, the client hypervisor will deliver apps to users on demand, no mess no fuss, and with true local performance. There'll be no two day on-boarding process to enable the contractor or consultant to connect his laptop to the corporate network.  Just turn up, plug-in, boot the corporate standard build and work.

In a future where mobility will have an increasing importance in all our lives, the ability to both support the contractor and mobile worker are where the game needs to change and where the client hypervisor will be the bigger win.


I agree w/ Brian. I think that client hypervisors will in most environments improve control, increase ease of management, allow rapid deployment and tighten security. The degree of impact each of these areas have on a given organization will vary, and will be of differening value. But overall this evolution in the way of doing things on desktops is a leap forward.


Brian I think it would be helpful to have a vote to understand what people may do. That vote would answer yes or no to the following questions.

1) I see a strong reason to buy Intel vPro (hardware lock in) enabled PCs/Netbooks/Latops etc when Citrix and VMWare annouce their shipping solutions?

2) Do you think you will buy Intel vPro enabled machines in 2010 to support the above?

3) Will you throw away your ThinClient investments to upgrade to Intel vPro?

4) Do you know Intel vPro has been hacked, check out network world?

5) Do you know how to patch Intel vPro?

6) Is Apple support important?

I hope you publish a version of this, and educate your readers on the hardware lock in strategy that is being played by Intel via Citrix and VMWare. We all need to wake up!!!!


Ah... AMD and Intel are part of the Distributed Management Task Force, Inc. (DMTF).



Im awake...



Certainly for standards, but not for implementations that lock you into one method of doing everything. MSFT is also an author of the OSD packaging format, did ltd things with it, created the MSI mess that came out of the Office team and now they are smoking SFT files, which will probably just become a form of VHD. VPro is still a custom Intel lock in that is increasing costs and is confusing the potential real value in the client virt martket. You are sleeping at the wheel if you think they give a hoot about DASH, it's about $$$$ since they are crapping their pants that they are loosing market share to thin client vendors. Just wait a year, and see how much VPro will matter when folks acutally have the stomack for client virt. VPro is 100% marketing with ltd value.


Let the market decide.


"We the customer" define the market, through knowledge, sharing information amongst ourselves, and our spend $$$. Unless we blindly follow the BS the vendors feed us who would love to lock us into markets they want us to believe in. That's why I don't have an avatar of a sheep, just kidding :-)



You saw the demo in Las vegas with the Sun "hypervisor".

that will enable you to do both in just two seconds!

Switch between machines on one client, and even take your sessions with you in an internet browser.

And they hide the clients hardware from the operating system. And as I demo'ed, this can be done in just two seconds.


@ Erik

Are you insinuating that SUN already have a client hypervisor similar to what is being touted by Citrix and VMware?  Or are you just saying that there is an extension onto VirtualBox?



Sun has a hypervisor that is on a gateway for over 12 Years. The only thing they did was never mention it. In earlier years it connected to Solaris only but nowday's we also use it to connect to Vmware RDP or DinamiQs DCP.