Do you think VMware is catching up to Citrix in the desktop space? Balderdash! Citrix needs to catch up to VMware.

Thought provoking post, Brian.  I am curious what your perspective is on what the mix will be between server-hosted desktops and “offline VDI” desktops will be once VDI+ is a reality.  I have a lot of respect for the vision that both Citrix and VMware are showing in this space.  However,  I think the one major factor that is holding back adoption of virtual desktops today is that it is starting with a server-centric approach and “offline VDI” is being viewed as a niche use case to put on the roadmap.  This ignores the fact that the overwhelming majority of PC desktops today are thick clients.   Add to this IDC and Gartner PC shipment data that shows that laptop shipments to businesses have eclipsed stationary desktop PCs, and it seems in some ways like stationary server-hosted desktops are the tail trying to wag the dog.  In a world where client hypervisor technology exists to support offline VDI, why wouldn’t organizations just start with that as their primary use case?  Relatively few benefits of desktop virtualization are derived from the fact that the virtual desktop is running on a server, and those that do (like security) can be addressed in other ways.  Also, by starting with distributed execution of virtual desktops on existing thick client devices, IT professionals can prove out the value of virtualizing desktops with their existing hardware footprint  without trying to push through a major data center investment in the current economic climate.  When a clear case is presented, by all means run virtual desktops on a server or even in the cloud.  However, I think making this step one in adoption of desktop virtualization is going to limit adoption of an extemely powerful technology.

Disclaimer:  I work for a company (Virtual Computer) that plays in the client hypervisor space.

I know that this article is a comparison of Citrix and VMware; I was thrown off by your idea that "Unless you have a very specific tactical need, DO NOT USE VDI TODAY!"  How do you feel about the Virtuozzo/Provision offering?  Seems like if Provision ever release the graphics acceleration they've been hyping, the only thing missing is the one-to-many disk streaming (which Provision also claims to have in development)... We're considering implementing the Virtuozzo/Parallels product, luckily we've already invested heavily in Citrix Provisioning Server to solve the one-to-many... Your article gives me pause, as you seem really firm on the idea that nobody at all has VDI quite right yet?

Good article Brian, though i have to say that we are back in the whole debate on ESX or XenServer thing ;) do you want a single platform of management or do you need the highend-features that VMware provides that differentiates.

I think it comes down to features needed in the company and price to pay for the product :) there will be room for both the technology lead and the overall delivery vendor with a good message ;-) you have to give Citrix that they came up with a cool message ;-)

Right now the battle is what hypervisor to choose, i do agree the next battle might be which VDI+ platform to choose. But i do not think that it will have to be the platform with the most genius technology that will win the battle :-)

Microsoft is still keeping fair market shares, even though most of their servers/services can be replaced by, in some peoples oppinions, better alternatives looking at them individually.

Rene Vester

Aaah, you are again trying to get us Citrix people agitated again :)

You well know that next year we will get our "user installed apps" and most probably advanced user profile management including streamed profiles, etc.

That is 3/5 for Citrix.

The question is what will happen with the offline and client hypervisor stuff. My bet is that in the long run Xen will be the main choice for "BIOS level" hypervisor in laptops and desktops (if this needs to be spelled out, the reason is device driver compatibility of Xen). However Citrix needs to cough up this XCI next year or it will be too late.

Brian,

With the Virtuozzo solution, users are allowed to install their own apps - unless you know something I don't?  VTZ has their whole container-based idea, so users can install whatever they want, as well as offering me the ability to create application templates to deploy on a larger scale... Has your experience taught you otherwise?  I have yet to get hands on, just quite a bit of thorough research...

Good post Brian and I would suggest the VMware announcements are fantastic for the whole VDI space and slowly but surely we are heading towards a real VDI solution offering but as we all agree it’s not there yet.

Nobody would argue we require the five technical capabilities that need to be added to today's VDI products but with each new announcement we are getting closer and whoever edges in front will gain momentum for now but the other players will surely catch up and then we will be on a level playing field again.

I’m just happy the VDI players are making progress to a real VDI solution offering but then again I would suggest VDI has a place in the enterprise as it currently stands.

It’s down to a good Integrator to map out requirements and what the current technology can currently deliver and then design the solution accordinary.

The who catches who debate will be around for some time but at least it’s pushing the relevant players towards what we all firmly believe in.  A true enterprise ready VDI solution offering.

Brian - I enjoy reading your posts but it looks like the goal for this one seems to be more to create a controversy (or as you may put it conversation) than it is to give the readers good valued advice :)

I put my thoughts down in a post:

community.citrix.com/.../5YcAAw

Folks - it is not about tactical or strategic or wall to wall. Haven't we all experienced this over and over again when it comes to client computing? We all get stuck in the technology architecture and we debate about it instead of thinking about the users, use cases and what value we are trying to obtain with the technologies that fit the user cases and user types?

Well - my post gives you a bit more insight into thinking use cases and I explain a bit more on Citrix's prospective on mobile and offline use case. It is a quick read:

community.citrix.com/.../5YcAAw

I'd like to see improved management tools in the requirements.  If you are going to dump the support of 1000s of desktops on administrators, at a minimum it needs to be as easy as managing a Citrix XenApp Session (if not easier).  One interface to find the vm information, running processes, events, alerts, hypervisor host use/assignment/statistics, provisioning server info, vDisk use/assignment/statistics, session stats, performance metrics, shadow access, etc.  I think this is a big obstacle for large scale deployments and just as important as those listed.

As far as number 5 user installed apps I'm really curious to know what organizaiton really wants to allow this? And how do you possible plan to achieve this on a locked shared image? Sure you could stream the apps out ala App-V after the image boots but the apps are still controlled by the datacenter. Anyways so that got me thinking and I suspect this is how the industry will go.

Consider this. Once the industry standardizes on a Virtual file format... and it will be VHD what is to stop citrix from offering a separate private mini VHD booted along with the managed image and available as a separate drive seen by the OS that users can write and save to install their own apps to? When they log off it can be compressed and saved off to their server documents folder like PST's are. It will happen when Windows 7 comes..

For clients I have worked with in APAC, the main business driver for VDI has not been in the office, it has been for remote users, as in people who need to work from home for most of the week, remote external companies who require access to applications that can not run in a TS environments where extending the network out to untrusted sites is not preferred, and this fixes that issue.

Citrix have their Access Gateway/Secure Gateway product that does this with minimum fuss without the requirement to open up a VPN Session (or you can, either or), perform end user scans to ensure the workstation is as expected and then receive a desktop.

I think comparing VMWare with Citrix product for product at the moment is just going to give you the perception that they are almost the same and deliver the same thing.  And in many ways they do in my opinion.

Just my 2 cents.

I love this funny game :

Mine is bigger than yours !

Both are good products with their strengths and weaknesses depending on your infrastructure you have and the goals you're after... Both will probably be better than the other if you include some infos from a  3/5 years vision (the info you have or you beleive in).

Rather stating than A is/will be better than B, you can probably focus more on in which case it is...

I don't want to buy/recommend the best technical product... I want the best solution for my company/customer.

Both are delivering the same type of services... One if probably better for WAN access... One is probably better for security of performance or price...

I know companies for which saving "manpower" with good administration tools is no use (and is much more a problem) as they can not fire them (like in public services)...

After that, I love to here from the vision information you can grab, the technical aspect of things but classification, hierarchy or "better-than-story" are definitly not what I'm expecting from your independant consulting company.

I think that looking at VDI future from a features list point of view will miss a point, rather we should look from end user (the "business" user) point of view and the enterprise (the "data center") point of views.

From this prespective I think we can say:

- Which platform is/will be more "data center

  ready":

 i.e. better management tools, better integration

 with current infrastructure, easier learning curve

 for the IT stuff , better migration pathe & co-

 exsitence etc.

- Which platform will help us improve the SLA

 with the end user ?

Looking from this prespective I do agree that none of those companies is fully ready for prime time.

Interesting assessment of what deserves a fail or a pass. Apart from the offline image support I would not have been so generous with the View offerings from VMware. For a start:

1) TCX is available for any connectivity between a Virtual Desktop and client utilising RDP or ICA. This includes any Wyse thin client running WTOS / Linux / Windows XP(e). So effectively if you are a Wyse client based implementation you can utilise any brokering solution.

2) PortICA is acknowledged as not on par as yet with the ICA stack experience provided by XenAPP but is improving, it is definitely an improvement over RDP. The Teridici offering from VMware is vaporware as yet so no judgement should provide it any weighting. In doing so you may as well also start talking about MS and Callista or Citrix Apollo (which at least is a downloadable product www.citrix.com/.../details.asp)    

3) The image management provided by View although a good step in the right direction still falls flat when you think about image management. How good is that system the first time you send out an OS or Office based service pack? Compare that to the use of ASIS from NetAPP which you can dedup the volume at any point in time allowing for build maintenance tasks involving disk space growth such as service packs to be consolidated back.

4) How flexible are the logical modelling options for any of the brokers (Citrix DDC or VMWare VDM) when thinking about flexible machine assignment on a global scale as a viable desktop alternative. This is an area where they find themselves wanting

I am a firm believer in VDI and have been involved for a number of years in it's delivery as a desktop replacement (not just an offshore or niche integration). With the technology being in such a fast moving development phase the way to get the best out of it is to be open and utilises the best of breed from everywhere including the hypervisor, clients, broker and storage.

Actually, on December 1st Provisioning server will have persistant differencing disks, so user installable apps is there.

And on not using VDI now?  Balderdash.

I have several clients that have no need for full multimedia redirection the version that is included with the "bastardized" Xen App as you seem to refer to it :) works fine for those staff.  Where we are seeing the biggest push for XenDesktop VDI is for replacing the XenApp Desktop Replacement that has been installed previously.

So how come you were referring only to future code for VMWare but not for Citrix products?

VMWare has no protocol story that is real today. Until they do, they will stay with the 10 customers they have for Desktops. If MS sticks them to RDP 7, then you are also stuck with Windows on the client only. Offline VDI does not mean checking machines in and out, that's just dumb, it's just too much. Better to go with alternatives like MEDV, Moka 5 etc

Intresting. Watch this video from 2006

www.pubforum.org/.../Day%2003%20-%20Brian%20Madden%20-%20The%20Future%20of%20SBC.wmv

Caveat - I work for VMware. There are a ton of customers doing VDI today. Heck, the first customers VMware had doing this were doing it in 2003 - well before the term VDI was even thought up. Citrix alone has been doing published apps for what, 15 years? If you're looking to do VDI then you really need to look at the use cases first. There are basically 3 categories of workers:

1) Work from work. These people don't need offline. They don't even need a really robust protocol. ICA vs RDP vs SPICE vs anything else really doesn't matter. The biggest thing you need for these people is ease of use, reliability, and a good bit of density on the back end.

2) Work from home. These people might need offline but typically not. Usually these are people like my wife who has a PC at home all the time and just needs to connect to the network and get her work image up. In the US these people also really don't care about protocol since over 90% of the US is on a high speed, low latency link according to the latest Nielsen NetRating survey. What you need here is a universal client that will work with anything as well as some method for handling remote printing, USB, and possible voice. All of that's available today from VMware and Citrix.

3) Work from anywhere. These are people more like me where I'm at home most of the time, sometimes in one of the VMware or EMC offices, and a lot of time in airplanes or trains completely disconnected. This ground needs offline. They need a good protocol since there's no telling what type of connection they'll have. They need a universal client they can use from a phone, PDA, laptop, etc. This is the group that should probably wait a little. They won't have to wait long. Check-in/Check-out will be here REAL soon. Decent protocols are here now. The rest will be here next year.

Nearly all of the cases for VDI I run up against are #1 or #2. Very rarely are people trying to tackle #3 since most of the time they have laptops anyways. So, does VDI work for most people today? You bet. If you decide to wait until 2010 like Brian says then your the guy I see on TV using stacks of dollar bills in the fireplace to keep warm. You're also probably one of those companies in line for a handout from the US federal government to bail out your poor business practices.

Mike, I agree with almost everything you're saying, with the small exception being "In the US these people also really don't care about protocol since over 90% of the US is on a high speed, low latency link"

Even with broadband Internet connections graphical content, especially flash, silverlight and anything 3D perfroms really poorly with RDP5 & RDP6, and I haven't seen anything that will change this in RDP7.  Even if you have a decent experience with a T1 speed connection, the amount of bandwidth that RDP5/6/7... consume is tremendous compared to Quest's EOP & Citrix ICA/Port-ICA.  So while RDP "might" cut the mustard for a single remote user where they are the only one on connection that is T1 speed at each end, when you plack a few users onto the connection the amount of bandwidth consumed by the content matters a lot.  On every test I've done RDP transmits 5-10 times more data over the wire than EOP and ICA do for anything graphical.

For everything else you said, I'm in complete agreement.  These things that Brian is mentioning are valid concerns, but not for a majority of business users.  If 50-75% of users can be satisfied by what exists "today", then worrying that we don't have everything for every users seems a little odd.   These things are coming sooner rather than later, so address the users that you can now, and leave the others for stage two.

It is great to identify the sore points, as it is unrealistic that a single technology that's been around for 2+ years can address every use case.

Brian,

You can cross off another one for VMware today. VMware View 3 was launched this morning and adds the offline component. www.vmware.com/.../overview.html

Patrick,

As always, EOP sounds great - it's just that we're going on month 3(ish) of hearing about it, and still no release date?  I know Quest keeps saying "Q3" - not much time left as far as that goes... Anything you're willing to offer?

EOP "currently" includes two components, i.e. Multimedia Redirection and Local Text Echo.  When vWorkspace 6.0 ships "this month" it will add RDP Graphics Acceleration and support for HP's RGS Protocol.  In 6.1 we anticipate adding GPU-backed OpenGL support, as this is currently in QA.

I don't ever "recall" saying that we'd ship the graphics acceleration stuff in Q3, as we got the first prototype in July, but I hear what you're saying.

If you are a Quest customer you can get the bits now from any of our SEs, otherwise you can wait until the general release of 6.0 later this month.

Brian - excellent article.  I love the way it has really stirred the pot among the many voices in the desktop virtualization market.

Like Patrick Rouse, I also agree with the way Mike DiPetrillo has segmented the virtual desktop market.  And Mike is savy enough to know not to atttempt to quantify the percentage of users actually fit into one bucket.  But I'd say bucket #3 the one Mike sits in and the one I sit in and my guess most people reading this comment sit in is far from a niche use-case.

The truth of the matter, as each year goes by, the Enterprise workforce becomes more and more mobile and personalization is not a nice-to-have but a requirement to keep a competitive, productive employee.  Anyone who has read Bill Gate's favorite yet always updating book "The World is Flat" by Thomas L. Friedman knows this is a reality of the modern world - personalization and mobility.

VMware View is a shining example of where the Desktop Virtualization market needs to go. And its pleasing as a VMware partner to see them embracing the same vision we have here at RingCube.

That said, this new function called "Client Virtualization with Offline Desktop" appears to check-in/check-out of the entire hosted VM to an ACE/Workstation/Player like client.  The problems with this “experimental use” feature, besides it not being GA, are the same problems client hosted hypervisor-based virtual desktops have always had:

1. Bloated size for storage and synchronization

2. Slower performance (at least 30% overhead)

3. Additional copy of the Windows OS

Could baremetal client hypervisor-based solutions overcome the challenges above? Yes, but there is an even-longer list of problems and costs with those types of solutions today and until an industry standard is reached, it’s safe to say we won’t see any significant (i.e. <1% of all Enterprise PCs) deployments of baremetal virtual desktops.

DOWNSIDES OF THIS HYPERVISOR-BASED OFFLINE APPROACH:

For the end user, the virtual desktop significantly underperforms compared to the native host which is why most users won’t accept the solution. Whether using quantitative benchmarks or qualitative analysis, almost any user would rather work natively on their Windows PC/laptop instead of working inside the hypervisor-based virtual desktop hosted on a Windows laptop or desktop.  The exceptions, of course, are users running Intel-based Macs and the ultra-rare Linux PCs. It's safe to say those users represent less than 5% of the enterprise workforce.

And these performance differences, despite some vendors attempts to hand-waive as no big deal, are not a matter of opinion but are supported and easily replicated with almost any 3rd PC benchmarking test - see here for some examples: www.ringcube.com/.../perf.pdf

For administrators, increased costs, licensing complexity, and legal risks are real issues. The check-in and check-out of a single copy of Windows across multiple pieces of hardware for each user has a significant financial impact in both cost and complexity.  Any company looking to use this “experimental use” feature without working out all your Microsoft license agreements is going to get hit with unforeseen increase in Microsoft licensing, potential legal fees, and/or a high likelihood of receiving a poor audit result due to failure to met software license agreements.

Are there better solutions for offline mobility, consistent end-user experience, personalization, and user installed applications while delivering full performance of CPU, Memory, I/O, and 3D graphics?  Absolutely, but looking only to the big 3 (MSFT, CTXS, VMW) will not be very promising today.  Instead, customers should seriously evaluate solutions from innovative, fast-moving startups in this space.

Shamelessly, I like the vDesk solution from RingCube especially when combined with a VDI-based offering. But there are a several other smaller vendors who are delivering amazing innovation right now who add richness to the desktop virtualization ecosystem and have formal partnerships with the big 3.

So delaying a virtual desktop project because VMware or Citrix might not have all the pieces today seems a bit short-sited and unnecessarily constrained.

Just my 2 cents...

Doug

@Doug,

First, are you the same Doug Dooley that I know from way back? If so shoot me an email to mikedatl at mac dot com.

I heard an amazing state yesterday at a conference hosted by SIMtone. I can't remember where they got the stat but they said only 28% of all laptops move from their original locations. I found that pretty staggering. Of course I look at my wife who has a laptop that never leaves her desk at home except maybe to move to the kitchen or the couch in the family room. If she was a VDI user then there's no need for check-in/out there.

I am interested to see why you say the performance is unacceptable for an end user when running in a VM locally on top of an OS. Sure 3D pretty much sucks if you're doing really high end work but that's gradually improving. Gaming is pretty acceptable for the vast majority of games out there. Heck, Sinbad runs VMware Fusion so he can play games and he likes it all right. Martha Steward does as well although I think they play different games. Games aside what are most of these mobile people doing? Word, PowerPoint, Excel, etc. After all these are corporate users we're still talking about, right? Those things perform just great in a VM. For every benchmark you show me where things don't perform I can show you one where they meet or beat physical. That's just the world of benchmarks. If performance was really so bad then why are there millions of people running Workstation and Fusion today and not complaining and trying to return it? Performance is fine if you really look at the use case.

Before I address licensing let me say you should always check with your MS rep before following anything that anyone says. There - legal check out of the way. We're talking check-in/out for VDI at this point. If you're running MS operating systems for the VDI session then the right way to license is with VECD (as crappy as that is). VECD means you've got a license for Vista on the accessing device. You also have SA and VECD on the accessing device. Vista is the key since Vista rights let you run up to 4 extra VMs of either XP or Vista on top of that same device. This means when you're running your session off the server you're covered by VECD. When you check out and run it locally you're using the virtualization rights of Vista. You're all covered under the same license. No extra costs besides just the horrible VECD licensing scheme (don't get me started on that).

Doug, you're a good guy (if you're the one I remember). Next time tell people you work for Ring Cube before using Ring Cube to say there are other alternatives and a Ring Cube performance paper to show poor performance. Just something I've been caught on before. People like full disclosure. Pretty sure everyone knows that I work for VMware by now but if  not then I do.

Good conversation here. Good to vet all of this out.

@ Mike D

Yes, same Doug Dooley. And yes, RingCube which is why I have it in listed my profile and when anyone clicks on my name it goes to the company website.

Sorry, I thought I disclosed it properly. I'll be more explicit.

Mike, I think we'll have to agree to disagree about the performance issue.  As you are fully aware, in the VDI world, deals are often won & lost because of the protocol differences of RDP vs. ICA.

Umm, that's PERFORMANCE and specifically the end-user experience.

vDesk on a PC vs. RDP vs. ICA - I'll take that Pepsi challenge all day.

Taking the latest server hardware with RDP/ICA against even a 2 year old Windows PC, I'm confident the locally rendered virtual desktop experience we provide will be much more responsive and preferred by end-users.

Heck, we should get Sinbad, Martha Stewart, your wife, you, and me and conduct a blind taste test of a RDP desktop, ICA desktop, and vDesk (locally rendered) desktop.

Again, I really like our chances on who delivers the best desktop experience to the user. Performance might not mean everything in the Enterprise but after 20+ years of people running locally rendered desktops to be asked to run a server/cloud executed desktop, Brian brings up a good point - why isn't 80% of the world doing hosted desktop and hosted apps today? An answer to that question is price/performance ;-)

PS> I'll send you an email now. Talk to you soon!

Doug

Today VMware released their new VDI product, called VMware View 3 . View is a complete solution, combining

I was talking to [my now colleague] Alex Barrett a few days ago about a story she was writing on VMware

Citrix's Robert Hammersmith took me to task last week for pigeon-holing XenDesktop as just an old

Pingback from  VDI: Citrix vs VMware « Calfo’s Blog

Another year has ended. We did a "year in review" article in 2005 and 2006 (not sure what happened