My prediction for mass client hypervisor adoption is 2013. Here’s why.

I can't possibly see client hypervisors becoming popular and adopted en masse any time before 2013.

I've written quite a bit about client hypervisors over the past few years. But now that we're getting closer to Citrix's production release of XenClient, and now that VMware is backing away from the client hypervisor, I'll take this opportunity to again put my money where my mouth is with a prediction: I can't possibly see client hypervisors becoming popular and adopted en masse any time before 2013.

Why client hypervisors?

I guess that we should take a step back for a minute and a look at why anyone would use a client hypervisor considering that I've written a lot about why they're NOT needed (here and here). Client hypervisors can actually be useful in a few cases:

  • Providing "offline VDI." Sure this is something we've been talking about for years, but the reality is that if you buy into the whole desktop virtualization thing, and if you spend a lot of time building images or whatever via a desktop virtualization environment like XenDesktop, View, or vWorkspace, it would be great if that same system could just be extended offline with a few clicks.
  • Providing "local VDI." I know this is technically the same thing as "offline VDI," but I love that we could get the single image management of VDI without the hassle of Provisioning Server or SANs. And I love/hope that it could be integrated with XenDesktop or whatever so that we didn't have to go with a third party product like Wanova (which works great today but which is something else that has to be bought, learned, deployed, and managed).
  • Security. Maybe it's my Bay Area affinity, but I really like what the whole TXT component of vPro can do in a client hypervisor environment. I love that a company could use TXT to ensure that a user's laptop only booted the proper hypervisor, and I love that that could be extended to ensure the user only runs approved VMs (or that different VMs could be certified to run with certain characteristics or privileges).
  • And of course, I love that the laptop manufactures will start to offer checkbox-type purchase options to enable the inclusion or activation of client hypervisors on the laptops they sell.

The client hypervisor adoption curve, 2010-2013

Here's how this is going to go down:

2010

Citrix will release XenClient as a production supported product. It will be fine. Not great, but fine. And even though companies like Virtual Computer and Virtual Bridges have been shipping stuff like this for awhile now, XenClient is the first thing that people will start to talk about. This is where the education will begin.

However, it will be very slight. Early results of our BrianMadden.com "State of the Desktop Virtualization Industry" customer survey show that only 17% of respondents even planned on using or evaluating client-based VMs by the end of 2011. And that's for all flavors of client VMs, not just client hypervisors. (If you haven't taken the survey yet, please do so now. It will take 8 minutes and help us learn more about the various industry trends.)

Will we see any success at all in 2010? I'm sure, but it will be very limited. Expect Citrix to tout the same few case studies again and again—case studies that are probably just really big pilots and proofs of concept with companies who traded marketing rights for license discounts.

2011

The problems with the initial versions of XenClient will be (a) limited hardware compatibility, and (b) limited integration with the rest of the XenDesktop product. But those problems will get worked out in 2011. Adoption will still be low though... A lot of lab testing and people learning, "wow, this is different."

2012

By this point we'll have a client hypervisor that's actually pretty usable and that will start to get some momentum of some big wins and huge deployments.

2013

This is when enough people will start to deploy client hypervisors, based on all the good things they heard about them the previous year. Certainly we're not talking about anywhere near 100% of users, but by this point client hypervisors will certainly be common enough that there's no risk of getting fired for trying them.

What about the hardware compatibility problem?

Ironically the limited hardware compatibility list (HCL) problem will sort of take care of itself as time goes on. This will happen from two angles, as (1) more new laptops replace old ones, and (2) the client hypervisors mature and embrace more types of laptops.

The bottom line is that all is not lost when it comes to client hypervisors. They'll still play a big part of our industry. But they just might not come on as fast or strong as we initially thought.

What do you think? Faster? Slower? Or am I missing the point entirely?

Join the conversation

31 comments

Send me notifications when other members comment.

Please create a username to comment.

I think one thing that may change your curve or impact it in a signifigant way that is not noted here is OEM deals.  A deal with a Dell or HP in 2011 could impact it and accelerate it a little bit. Now this doesnt mean it will be used, just it may be be "There" though not really noticed (kinda like the BIOS is today). Then you get crazy numbers like "total 1 millions" client side hypervisors out there, but maybe 5 of them actually being leveraged.


another potentially interesting story would be a use case for non-business users. Kind of like looking at the iphone... not REALLY built for business users but so many uses/tools and so much consumer demand businesses are pulled to support it. Now the average user has no use for a hypervisor so this is a LOOOONG shot. but, you never know.


Cancel

Seems like a pretty safe estimate.  Most new business technologies take a few years to move from PoC to pilot to mass production. The client hypervisor is just another flavor of virtual desktops and it will be used by a percentage of the user population, just like the hosted vdi model and every other model that is out there.  They all will have their place and time.  


You also have to wonder if the BYOPC, for organizations that build into this type of program, will be the early adopters of the client hypervisor technology.


Cancel

The biggest problem is due to Intel investing in this technology with Citrix they are still insisting on vPro. TXT has been hacked news.softpedia.com/.../Intel-vPro-Hacked-101286.shtml last year so it's not as secure as you may think. I'd rather see full disk encryption type technology native to the hypervisor. As this thing scales, vendors need to remove the TXT requirements. I see nothing in the XC architecture that requires this. Intel in effect is trying to raise the costs of laptops by forcing vPro. They need to make this free and in every laptop including consumer to make it real. This will not matter for the early adopters but long term this remains a barrier.


We should also understand that MS does not want this to happen. They will kick and scream like school girls and do everything in their power to slow this market down, poo poo the use cases to some corner use case, bully the OEMS and Citrix until they have something in Window 8 which of course will suck like all their 1.0 products that are three-five years behind the market. This is probably the only technology that the MS client team cares about in the realm of desktop virtualization as this affords them less control of the client with Windows. This is also the key reason why Type 1 client hypervisors will become relevant and in time MS will be forced to move. Today it is clear to me at least that MS does not care about this technology and does not believe in the BYOC/D type use cases either which are not exclusive to client hypervisors.


The other problem will be management. The HCL list, OEM adoption, usability and stability etc will get better, and we may start to see this become a more prevalent architecture by 2013, but management will still be a huge gap. MS will try to sell SCCM which is on such long release cycles it will take years. Startups like Unidesk/Wanova will help seed the market, but until the big guys move via build or buy adoption this will remain nascent for the masses.


Certainly basic management capabilities will exist to do basic management for the hypervisor with some things like patching, but it will be very much IMO still be done as a distributed management model at the guest level for some time to come using existing tools making laptops even more complex to manage. That may be fine for early use cases in 2013, but the centralized management story of Windows at scale on distributed client hardware with awesome multi guest usability is still 2015-2017 IMHO.


As a result I believe hosted desktop virtualization is far more relevant for years to come to get any real business benefits. Client side is a 2015 target in my mind for broad adoption with management at scale.


Cancel

I have to agree with Ron that the OEM deals are key.  The HCL issue will be solved on all business class laptop sold as the manufacturer will include the relevant drivers.


Laptop manufacturers are always looking for a way to differentiate from the competition, before this it has been by installing bloatware on the OS at manufacture.  The client hypervisor changes this.


It provides manufacturers with a huge opportunity to provide additional functionality.  By running a tiny linux VM behind the Main OS, they could provide external virus protection (not available yet, but definitely coming), firewall, restore, upgrade, KVM support, etc. etc. etc.


IDC say Laptops will make up 70% of all PC's sold by 2012.  That's a hell of a big market to try and grab a bigger share of.


If we are just thinking about it from a desktop virtualisation point of view I think 2013 is too optimistic, but when you bring in the OEM advantages, maybe it's too conservative.


Also for the enterprise, it's a new way to deploy and control images.  Ever used PXE? It's rubbish!  It's outdated and IMHO not fit for purpose.  Client hypervisors supplied with business laptops will solve this.


Cancel

I’ll stop just short “missing the point entirely,” because I know you are not, but here are some thoughts to consider:


1. I don’t really see any description whatsoever in your analysis as to why client hypervisors could not achieve critical mass sooner, aside from charting the development of the whole product segment against a single vendor that, while certainly relevant, is not the product maturity and innovation leader right now.


2. If I read into this piece, as well as past posts, I get the sense that you believe that the client hypervisor idea is too big for a small company to execute.  Time will tell whether this is true, but I encourage you to give some thought to how a couple of major partnership wins can change the market visibility and trajectory of a small company with superior technology instantly.  There are a lot of large and influential companies out there who are not ready to concede this market to Citrix.


3. On the surface, someone could look at your post and view it as discouraging to near-term adoption of client hypervisors.  I think it is important to highlight that the point of mass adoption is not the first point at which a market segment is successful and delivering significant value to users.   According to various reports I have seen, there are over a billion PCs in use, with a million new ones added every day.  This market is big enough to generate many, many IT success stories (and multiple successful vendors) before we even hit 1% market penetration.    If client hypervisors do truly achieve mass adoption in this market by 2013, this would actually be an unprecedented technology shift that would dwarf anything else that has happened in the virtualization space, including server virtualization, by orders of magnitude.


4. There is already at least one client hypervisor that is “actually pretty usable.”  I would love to see some type of objective product maturity and readiness criteria applied to all of the products in this segment.  I have seen you guys do this type of coverage of other product segments, but it seems like every piece on client hypervisors is high on sensationalism and large vendor flag-waving and short on true analysis of technical merits.  I am sure that mainstream analysts like Gartner/Burton Group will get to this eventually as an extension of the great work they have done in this area with server virtualization and hosted virtual desktops.  However, I think folks like you and Gabe are important voices from whom your readers would appreciate more substance and depth on this topic.


Doug Lane


Virtual Computer, Inc.


Cancel

Great article. I agree with the 3 year mark, client side virtualization needs a lot of time to mature to ensure proper management and flexibility is offered.


@Ron - You said it. OEMs play a very large part in this, because being able to purchase laptops/desktops with a client hypervisor pre-loaded will not only make the hypervisor ubiquotous, but will also make the transition from Traditional endpoints to VDI endpoints much smoother and easier if/when that time comes.


Client side computing is just as important as Server side, so there needs to be mature offerings of both that dynamically work together to create a constant flow.


Cancel

@Doug


Much respect goes out to you.


I understand where you are coming from. Being the one vendor that offers the richest client virtualization technologies to date that make you years ahead of the competition should raise at least one eyebrow on this article.


With reading your post on the topic I do agree that the lack of information of this article that you have thoroughly given in your points shouldn't go unnoticed.


The reality is that client hypervisors today are ahead of it's time, well actually to correctly put it the time for integration is now.


The points Brain make are valid.


- client hardware integration needs maturity (the new hardware should be looked at more than the old)


- SBC(VDI) vendor integration needs maturity (dynamic workflow)


- Storage considerations?


- Problems that haven't been found or discussed?


Seeing that it took several years to get one vendor's product to acheive Enterprise Ready SHVD status from Gartner, I would have to assume a similar timeline for Local Hosted Virtual Desktops (LHVD) given the areas it needs improving.


3 years can be seen as too long or too short, but the bottom line is that maturity still needs to occur for the larger picture.


Cancel

App detective put the MS shoe on top of the OEM arguement that I didnt even think about. I think the OEM piece will be key but really the question will become HOW MUCH pressure will MS put on Dell, HP, etc to NOT OEM a client hypervisor, or delay it, or minimize the number of laptops/desktops to support etc.


As for the disk encryption... Its a big deal but maybe could be handled another way and not in the hypervisor :-)


Cancel

@doug The challenge I think you guys face is that you have chosen to focus on management out of the gate which squarely puts you in competition with MS and SCCM. You will meet lot's of resistance there from MS and the old school IT people who are wedded to that POS. While I admire you for that and hope you get lot's of traction, I think the market today is still not convinced about the maturity of the Client hypervisor itself and usability of multiple guests. Therefore I think Citrix are smart to focus on that (it's technically hard) and let the management thing play out. They will take time to get that to a usable state. As a buyer I am looking for hypervisor stability first and then management will come. When it arrives I don't want the hypervisor vendor to lock me into his/her management stack either. I would like to see a management eco system form around the most stable high quality open hypervisor. I'd also like to see better integration with the hosted desktop infrastructure. That is not something MS is going to do, and hence I don't give a hoot what they do in time and 100% I am not marrying SCCM.  Hence why I believe XenClient is such a big deal. Over time I certainly believe you can add value on top of XC as a management provider that is part of the eco system. Again just my opinion, as a buyer I would be more reluctant to trust you with the hypervisor given the resources that it takes to get it right over time. I understand that as startup you will tell me you are great, but startup delusion is also something I spend my life guarding against, case and point is Atlantis computing.


Therefore my direct feedback to you would be, tell me how your story will evolve with XC? Are you competing at the hypervisor and mgmt level? That may be current state, but what are your thoughts looking forward? To me if you are competing at both levels, it's a much bigger leap of faith that I am unlikely to make broadly, which is why I think all this stuff is going to take longer than we would wish.


Not trying to discourage you, I hope to see you get traction but wanted to express my concerns as a buyer. Good luck!


Cancel

@Ron, if MS bullies the OEMs with Windows to slow innovation, kill competition by marrying Windows to the client hypervisor then I think the European Union should give them a call and remind them of the past.


This is why I think it is so tragic that VMware appears to have pulled out of this market. MS will push Citrix around, but VMware have the mandate to screw MS, which we should all support as it will lead to innovation. Are you awake Mr. Maritz?


Cancel

@Icelus - thanks for the props.


OK, so far we have:


- Industry forces such as MSFT, pace of OEM adoption, etc: Totally legit point.


- Disk encryption: Solved. Included in NxTop Engine, along with remote wipe, time-based expiration/lockout, peripheral filtering, and a bunch of other security goodies.


- HCL: Solved.  NxTop Engine runs on both Intel vPro / VT-d systems and older VT-x systems.  Support today in GA product for Intel and NVIDIA graphics.  Beta available with AMD processor and ATI graphics support.  What's missing?


Cancel

Regarding OEM integration, that's a good point that I meant to make but I forgot. I agree we'll see that almost right away, but I wonder how many people will actually select that option, even if it's free? If I'm a customer who's thinking about client hypervisors in the future, maybe that would help me decided which laptop I buy. But why would I select the "yes" option at purchase time when I could just install XenClient later on?


My point is that broad OEM support will be crucial, but just the fact that it's a checkbox at purchase time won't matter as much.


Cancel

@doug how do you deal with usability across multiple guests. Hornets nest of problems to solve.


@Brian. It's about the drivers more than anything else. If the OEMs pick hypervisor X to test against then we will get better support to make sure stuff works. The Hypervisor vendors can't scale to make this happen. This is where I predict MS will leverage the OEMs to build support for their solution which will be a late entry to a market once proven like everything else MS. No innovation, just play catchup and feed bottom up and erode other's market share.


So as OEM just ship these things at high quality, the checkbox and who installs it won't matter as you point out. What it will do is increase the confidence of the buyer that the hardware features will work.


Cancel

@appdetective - Thanks for the feedback.  I think it is fair point about our management strategy, though we do try to strike a balance between not being afraid to ruffle feathers and extending the olive branch to any and all existing ecosystem players to see how we might complement what they are doing with customers today.  


With respect to our client hypervisor strategy, we are very committed to having our own for the forseeable future.  I also think that if you took management out of the equation and just did a straight client hypervisor bake-off across a comprehensive list of areas of importance (user experience, HCL, breadth of peripheral support, stability, etc.), we would come out on top.


If another client hypervisor option emerges that customers see as having advantages over ours for their use case, we will definitely take a look at supporting a heterogeneous mix of our hypervisor and others, as is common practice in the server virtualization management space.  However, our experience to date has been that when prospective customers try NxTop and try XenClient, the conversation is not ending with them asking us to manage XenClient.  That may very well change someday as XenClient matures, but in the meantime, we think our time is better spent making our stuff better.


In terms of your question about user experience across multiple VMs, we have invested quite a bit of effort into the development of paravirtualized drivers for all major I/O areas like disk, network, USB, mouse, etc.  If performance trade-offs need to be made between multiple concurrent VMs, we give precedence to the VM in the forefront, which in most cases makes any performance hit to other VMs invisible to the user.


Our performance prioritization methodology also limits the impact of any background management tasks we are doing.  For example, we'll try to use idle time to move bits over the wire, collapse VHDs locally for performance, etc., but if detect an increase in user activity we will dial our stuff back and give precedence to user experience.  Not sure whether that fully answers you question, but hopefully it gives you a flavor.


Cancel

Brian-


I think one way to make someone "check that box" is to offer OS options that MAY not be available otherwise. Now dont get me wrong, the vast majority of users dont "need' a second VM. But picture college kids that could check the box for the hypervisor and get both the Windows OS and a preconfigured Linux OS or multiple VMs. They could run both, or bounce back and forth, etc. Of course that is for those so inclined.


But it could also open the market for "Check this box" and get a standard Windows OS and a Security appliance (that will do IDS/IPS, AV and firewall in band) and an appliance that will allow for file sharing and what not (outside your secure environment) etc, etc.


So there may be reasons to put it out there and offer some consumers some products that we havent even thought of yet... And if Dell had such a product vs HP or HP vs Dell...


I think there are some real options and inovation that we have yet to see. MS will not like it much, but if they REALLY wanted to get ahead of it or in on it, they could.  


App Detective-


I am not sure what is going on at VMW with this. They officcialy say they dont see a market and customers arent begging for this product. But I think that depending on the conversation and the framing you could get a business customer to say that. But I also believe that without the secondary VMs/products that this would enable most cant envision what they would do with a client hypervisor beyond some small uses cases that everyone talks about.


Its almost like the PC.itself. The numebr of uses we have for the PC today could not be articulated or even envisioned by most early on. The PC was a 'business thing' except for a few people who had the vision and understanding of the doors that would open and surprises behind the doors that they couldnt think of.


I really think the client side hypervisor is a "build it and they will come" model.


Cancel

@Ron really good point on what other service VMs could do. I know XC enables services VM's as part of their architecture, and I guess that is also what McAfee is also doing in their VDI offering. The service VM architecture I agree could really spawn some innovation. As for VMware. I got a copy of a presentation the other day, with VMware talking about the future desktop as a cloud that is available today and picture of the dudes grandmother. From what I could tell the presentation tried to imply that the it's cloud based. Yeah sure one day for a subset of users, but nothing on solving real problems. More importantly I think that is their strategy, endpoints that connect you to their services. This to me just further shows how they are irrelevant in the desktop and don't understand the real pain points and just BS about futures that may never mature.


@Doug ok got it fair enough, although I promise your hypervisor will add no value over time as it be commodity and open source. I'll have to check into your claims at some point, but for now I'll let it be. That said regarding service VM's, does your architecture allow other to write secure service VMs. I know Neocleus (not sure if they are still in business, Brian call them please )was going to do this and they touted that as a key differentiator vs. you guys.


Cancel

@appdetective - I agree with you that the client hypervisor will be a commodity eventually.  But right now it is a differentiator.  Strange business. :)


It's a calculated gamble spreading our focus between management and the base hypervisor, but it keeps us in control of our own destiny and even as a commodity the hypervisor adds value to our company.


We have not traditionally offered a services VM.  However, this concept is being added in our next release (currently in beta).  It is not yet an open services VM that anyone can develop for.  To start, we will control what can be included, with the admin determining which, if any, of these elements is published to the user's machine.


Initially, we are including some end-user "goodies" like a web brower and Skype client, as well as the ability to run a remote display client in the services VM to isolate server-based application or desktop sessions from other VMs of varying security that might be running locally.


Cancel

@Doug


Some great feedback from the readers!


I would like to offer the 20 million dollar feature:


The ability to put the client hypervisor into RAM.


RAM mode is a key to stateless computing initiatives across many sectors.(no pun intended).


Let me know if you want more info.


Cancel

It's my turn to put my money where my mouth is and say that "if you do build it, no one is going to give a crap about it." It won't happen in 2010, 2011, 2012, or 2013. That's all I can say for now, but I'm sure we'll circle back to this topic in the future. Companies like Virtual Computer and Neocleus might not even be around in the next 12 months.


Cancel

@edgeseeker...


In defence of Virtual Computer....


The time line predictions for the masses are possibly accurate but I know that Virtual Computer have quite large deployments out there. I also know a organizations waiting for the next version before raising their purchase orders.


It's also interesting to see what the outcome of Quest's and Virtual Computer's blossoming relationship may bring in the future. - even if its only the vWorkspace client at the hypervisor level - that adds tones of value in my place.


We're aggressively evaluating NxTop and have been impressed with the flexible management architecture and each version of the hypervisor is noticeably better than the last.


Like Quest, Virtual Computer listen to the customer (or prospective) - not dictate.


I'll leave you alone now as I have some respect for you being a vWorkspace customer - even if its with that container crap ;)


Dan.


Cancel

Brian, thanks for the mention. I am late to the party here as I just got back to town late last night.


The integrated off-line VDI is by far one of the most exciting features customers are asking for. That and the ability to handle remote branches.


I think this discussion should add a different dimension - why are we talking about stand-alone client-side hypervisors? I think the adoption is much more dependent upon the ability of vendors to offer a complete desktop management solution that integrates VDI, Off-line VDI and remote branch support.


I just spoke to a large institution in Boston yesterday that was frustrated with trying to cobble together solutions from multiple vendors and was very excited about getting all the pieces from one vendor.


That is what we are seeing from our largest customers worldwide and that is what I think will be the key factor in CSH proliferation.


My $.02


Jim


Cancel

@Daniel


My container crap is a lot faster, scalable, and sensical than your bloated hypervisor crap.  Sounds to me as though you have a personal thing against Parallels. But I won't rush to their defense as their technology stands on its own.


In defense of Virtual Computer? You make it sound as though I waged an attack on them. No so. I'm simply making a prediction of my own that client hypervisors have little to no merit, and what client hypervisor companies have set out to accomplish can be done in other overlooked ways.  


I've repeatedly said on this blog that Microsoft will not allow anything to come in between it and the PC hardware.  Therefore, I expect all the client hypervisor initiatives to eventually fall flat on their faces.


I too will leave you alone, but will circle back at you in 2011, 2012, and 2013.


Cancel

@edgeseeker...


I see what you did on that last line.


What are you basing little to no merit on?


Cancel

@Daniel


On some stealth technology I saw recently.


Cancel

@edgeseeker


With the announcement of Intel buying McAfee for $7.68b so they can integrate security into their chips how do you think they will deliver that?


In a VM service ontop of a client hypervisor maybe?


How will I be able to leverage the future of McIntel's virtualization-aware secure chips if I used containers?


Will I be able to?


Cancel

@Icelus


This acquisition is the most exciting one that I've seen in years, but it also stands an equal chance of being the dullest.


If Intel is merely acquiring McAfee for the sake of getting a bigger piece of the "corporate spending" pie, then this acquisition wouldn't be game-changing in my opinion. Intel is loathed by Wall Street because the growth story has been inexistent for quite some time. This acquisition could simply be a move to dispel this stigma. This could be a big Cloud Computing play.


On the other hand, if this acquisition is motivated by a higher vision, you should then expect a whole new generation of chipsets to emerge, with security and application awarness at the very core. This could also be a very big Cloud Computing play.


Of course, I would expect all platforms to be able to leverage these anticipated chipset functionalities, i.e.,  Windows, Linux, hypervisors, containers, etc. I'm confident such expanded chipsets would deliver the new capabilities to the upper layer as transparently as today's BIOS/firmware.


This is all speculation, though. I still think this acquisition is motivated by the need to satiate the lack-of-growth story. (At least, short term).


Cancel

@edgeseeker


good points. I think Citrix gave Intel an innovative push recently when approaching them with the Project Independance partnership (aka XenClient). I believe that may have triggered them to say, we need to do more to deliver security to desktops than just provide hardware. I am willing to bet service VMs are on the horizon.


Don't dissapoint me and switch over to the hypervisor side anytime soon. I learn a lot from your viewpoints on your pro-container posts.


Cancel

@Icelus


I'm sure desktops have something to do with it, albeit very tiny :)


I'm not married to containers, per se. I just think it's the most sensical solution for desktops, at least in my environment. I'm also a heavy user of hypervisors, both ESX and Hyper-V.


Cancel

@icelus, service VMs is what I think what will set XC apart from virtual computer etc. The McAfee of the world are far more likely to build for XC or other large vendors not a small vendor.


Cancel

Does anyone ever wake up and think ... google apps, plus drop box, plus receiver (origami), plus sso cloud authentication, add iPad and Done?


Cancel

@t.rex


Never. I usually limit myself to thinking about coffee.  


Seriously though, I might consider that as the primary toolkit in 2-3 yrs, but I'm sure I'll still be running Windows apps locally for some tasks.  


Visio as a Google App sounds as though it would not be a good move.


Cancel

-ADS BY GOOGLE

SearchVirtualDesktop

SearchEnterpriseDesktop

SearchServerVirtualization

SearchVMware

Close