Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate - Brian Madden - BrianMadden.com
Brian Madden Logo
Your independent source for desktop virtualization, consumerization, and enterprise mobility management.
Brian Madden's Blog

Past Articles

Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate

Written on Aug 12 2009 18,190 views, 43 comments


by Brian Madden

I’ve tried (unsuccessfully) over the years to introduce new industry terms. (xASP, xDI, farmlet, VDI+... The list goes on.) Even though I’ve failed repeatedly, I’d like to try again by introducing “Madden’s Paradox,” which describes the scenario where you decide to choose VDI over TS, but then when you try to make your VDI environment manageable, you end up removing the functionality that was the reason you chose it over TS in the fist place!

Let’s explore this. First, our assumptions:

  • “VDI” means connecting via a remote display protocol to single-user desktop OS instances running as VMs on a remote host.
  • “TS” means terminal server, (or what Microsoft is now calling “session-based remote desktops”), where a client uses a remote display protocol to run a desktop as a session on a multi-tenant Windows server.
  • “Server-based computing” is the concept of using a remote display protocol, and describes the method of access for both VDI and TS.

Okay, so VDI and TS are both server-based computing. This means that as you’re designing your environment, you first decide which applications and/or users need server-based computing, and then you decide which flavor of server-based computing you want: TS or VDI. Since TS is cheaper, it makes sense that you’d always choose TS except in the cases where you have a specific business requirement that can only be solved via VDI. I’ve written ad nauseam about why you’d choose VDI over TS, but one of the reasons that VDI lovers keep on talking about is the fact that users run a real Windows desktop OS that they can personalize in ways that can’t be done with Terminal Server. Fair enough.

However, the very next slide in most of these Pro-VDI presentations then talks about how you need some kind of storage management system that lets many VDI instances share a single master image. (Citrix Provisioning Server, VMware View Composer with Linked Clones, NetApp Flex Clones, etc.) The problem with using all these products in VDI environments is that you essentially turn your VDI back into a Terminal Server (in terms of all users sharing the same image), essentially killing the reason you were using VDI in the first place! Every single dynamic personalization technique that exists for VDI also works on Terminal Server. (User Profile Manager, User Data Disks, Group Policy Folder Redirection, third-party tools from AppSenseRTOREStriCeratScense.

So that’s the paradox: You only choose VDI because you want server-based computing but TS can’t cut it, and then to manage your VDI environment, you implement a shared master image system that makes your VDI behave just like Terminal Server.

Exceptions

Of course this paradox doesn’t exist for every single case. There are other reasons people choose VDI, although I still argue that many of its advantages are lost now that Terminal Server has per-session IP addresses, a multi-session installer service, and CPU management. Oh, and don’t forget that App-V or whatever app virtualization package you want to use works just as well on Terminal Server as it does on VDI, so that’s also not a reason to go to VDI. My point is this paradox only applies if you’re using VDI because it’s easier for users to install their own apps and own their own environment.

With that, let’s continue our democratic ultimate peer review process. Please share your thoughts about Madden’s Paradox. Am I right on or am I missing the point?

 
 




Our Books


Comments

Jim Moyle wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 4:29 AM Link To This Comment

Brian,

I absolutely agree unless you aim to fulfill the potential of VDI to enable full user personalisation, including user installed applications.  You end up just recreating a terminal server environment except with extra licensing costs.

In fact I recently wrote a blog article on exactly this subject.

www.jimmoyle.com/.../why-is-vdi-changing-into-terminal-server

Clayton Price wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 5:39 AM Link To This Comment

Citrix were on the money approx 10 years ago when they said "It's all about the apps".

In my opinion, streaming apps to both client targets and terminal servers is the biggest advance in this/our space.

VDI "hype" is simply the result of a virtualisation company workshop where someone said "We've won the server market...what else can we do with this sh!t?"

Why virtualise the OS? Spend your time and resources on a sound application delivery strategy, your profile management and your end user device management.

Hell, you're still going to need all of these things if you go VDI or not !!!

Scott Holmes wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 6:20 AM Link To This Comment

Hi Brian, great article.

Following up on your Parallels post the other day, I thought I'd try and communicate where our VDI offering fits into the 'Madden Paradox'...

As you know our 'Virtuozzo Containers' technology uses a delta concept similar to linked clones, with the key difference being that the delta's are server-side, not storage side, and consume very little resource even when running. When thinking in terms of the Madden Paradox, the key criteria is the storage requirement of a 'Container', which for the OS components needed to execute typically requires only about 100MB per user. Also, when not in use there is zero utilisation of CPU, memory etc (only disk is still consumed).

Which leads to the big difference when compared to the SAN delta concept... We can leave the users Container in existence at all times, without an enormous storage cost!

This delivers 2 major benefits when compared to the hypervisor offerings...

1) Users can customise the environment (within the confines of Group Policy) in the same way they could on a local OS, and the changes are saved.

2) The broker can be configured for persistent logons, meaning users get the very same STATE when they connect from different places - documents can be left open and returned to later, etc

All of this comes at a minimal storage cost and, in conjuction with our bundled broker (Quest vWorkspace), the application virtualisation/streaming platforms and the user profile abstraction offerings, Parallels VDI can deliver all the other features you would get on a hypervisor VDI platform.

One other big kicker... The Microsoft VECD licenses don't apply to us! We use Microsoft Datacenter licensing plus TS CAL's, which delivers huge savings from a MS perspective and further enhances the ROI proposition (not to mention gets you out of a subscription cycle and leaves OS upgrade expenditure up to you again)

So to summarise, I agree that at present there are many compromises being made in the SBC world and people are struggling to find the real value. I'd like to propose that Parallels VDI offers a functional and economical hybrid of TS and VDI architectures, and if anyone has any questions or comments I'd be happy to discuss further!

Regards,

Scott Holmes

Solutions Consultant, UK Enterprise Division

Parallels

appdetective wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 6:59 AM Link To This Comment

All this broker nonsense means like TS you are adding a layer of complexity into the Desktop environment for no good reason. Sorry to repeat much of my previous post. RDP to a desktop top does not require a broker, I can just connect. I want ICA/HDX directly to a Desktop for high tier users. For those folks they will not share a zillion VM's per host and it wiill be done on local disk. Shared Disk is too expensive and increases complexity and introduces single points of failure. It's a all about reliability for this tier of users.

For costs the brokers suck today and will take time to mature. TS is the best broker today. So PLEASE tiers of desktops please, high tier not cost driven and the rest that will take time to evolve TCO on a desktop OS, so go with TS today if you want it.

So sorry Brian, fail again :-) Go explain the use cases to the world and have the vendors understand that we want tiers of Desktops including the option to just connect to avoid all the BS in the middle that we don't want for a class of user.

Daniel Feller wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 8:02 AM Link To This Comment

Brian

You say TS solves all of the problems (or most of them), my question is what about the device that these TS users are using to get to TS?

Some will use thin clients, yes. That is an embedded OS. Why not provision the embedded OS to the thin client or desktop appliance?

Some will use their old PC repurposed as a thin client, yes. Why not use VDI to provision that repurposed PC?  Then you can manage it centrally.  

What about the users with newer PCs? We want to use the local hardware don't we?  If so, then we stream down apps from TS, but we still must manage that PC dont' we? Well, again, provision down the OS.  

My problem when people talk about VDI is they are always focusing on the hosted virtual desktop model. In my discussions, VDI is so much more. It focuses on desktop virtualization across all types of desktops to be able to simplify management of the end point.  If you don't have an endpoint, how do you expect to see and use the application?  You don't.  

Daniel - Lead Architect - Citrix

(twitter @djfeller)

Levi Beckner wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 9:13 AM Link To This Comment

Today it might not matter.  I don't think the magic of VDI has been brought out yet.

Today I think bandwidth makes Madden's Paradox pretty much true.  But with more bandwidth, VDI will have more potential than TS.  I think there will be more options for managing the desktop.  That is, protecting the user community from itself while allowing them more freedom.

Claudio Rodrigues wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 10:37 AM Link To This Comment

And just a quick comment. Tim Mangan can definitely step in here. At least for the previous App-V/SoftGrid clients the behavior was NOT the same under TS when compared to an XP client for example. Microsoft was even aware of that when we had a high profile customer trying to move pretty much all its apps to App-V to be deployed on their TSs. The end result, considering the load they had and the packages (they included Office full on some), when on the TS was not good. Microsoft at the time determined the issue to be on the way App-V worked on a TS and for that reason it would not be possible, under that load, with that many users and with high logon/logoff ratios, to get it to work properly.

That issue was not there when using XP with App-V. The multi-user aspect of TS in that particular scenario, introduced issues with App-V. That was some time ago and maybe 4.5 or 4.6 got that sorted out and now we may be able TS/XP App-V clients do behave the same exact way today. It was NOT the case before for sure.

caustic386 wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 10:49 AM Link To This Comment

I hate to side with the sales rep, but in terms of what exists TODAY, Parallels seems to be the only game in town.  When I watch the TS vs. VDI debates, it really does seem like VDI doesn't stand a chance as long as people come at it from the hypervisor perspective.

Citrix, VMware, and MS all seem guilty of just taping together a few things they had lying around to try and come up with a VDI solution while the term is still trendy, but to get a comparable experience out of their offerings, the IT staff has to do at least 50% MORE work.

As far as I can tell, with what exists right now, Quest/Parallels on the server side, and Citrix Provisioning Server on the end points is the only possibility of making things happen, and that's assuming you repurpose old PCs.  With HP or Wyse selling thin clients for $400, plus that super-duper experience of dealing with XPe, why would anyone bother?  With any hypervisor solution, you'll hit $1000/user easy before you get anything usable, and even then, then users (and probably IT staff) won't like it all that much.

appdetective wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 11:12 AM Link To This Comment

@caustic386 your challenge is you don't get the value of why and are making the 101 mistake of taking a cost angle. Sorry no offense

Claudio Rodrigues wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 11:47 AM Link To This Comment

@appdetective With a sinking economy, how can the upfront implementation costs be ignored? I 100% agree that cost is not the only thing to be considered BUT it is one of the factors for sure. Depending on the company this may be the driving factor. But again, when discussing this with Harry Labana at BriForum, the cost aspect of VDI as of today is something that cannot be simply ignored.

I get the value of VDI and the fact there are several cases where it would make sense for sure. But there are several road blocks, technological or economical, that prevent, as of today, the mass adoption of VDI. Several of these I pointed out on my blog at http://www.wtslabs.com/blog. Cheers!

Christoph Wegener wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 11:50 AM Link To This Comment

Brian,

I understand you make your assumption that VDI = HVD to be able to demonstrate your point.

However, the VDI = HVD assumption does not fully represent Desktop Virtualization.

There are a lot of technologies that comprise desktop virtualization and many of those technologies overlap with SBC/TS technologies (user personalization, application virtualization).

But the future goal for desktop virtualization should be what Gartner calls the composite work space.

www.gartner.com/DisplayDocument

And a lot of companies, that you cover on your website, are working on technologies to make the composite work space happen.

Right now, the redundancy between TS and VDI might still be very real.

But I hope that this will change very soon. And it won't be Citrix or VMware or Microsoft who push the composite work space. It will be the smaller, more innovative companies like Unidesk, Neocleus, Teradici and MokaFive!

Christoph

Tim Mangan wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 12:56 PM Link To This Comment

Brian,

Are you saying the VDI is the new bait-and-switch from the industry to pump up revienue?  

caustic386 wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 2:42 PM Link To This Comment

@appdetective - I'm happy to listen to the "value of why" if you're kind enough to explain it?

Today's hypervisor-based VDI = more work, and more $ (task-based users aside) than a standard desktop

Did I miss something, specifically?

Adam Wilson wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 2:43 PM Link To This Comment

I think Claudio touched on a good point.  He mentioned that App-V worked in an XP environment but not in TS for a particular group.  As TS administrators we've become very good at making applications work in multi user environments even if they weren't designed to do so.  It takes a very specialized skill set to make the stars align to get some apps to play nicely together.  Also,  many of the apps our organization uses just don't virtualize now.  I think VDI takes away some of the need for that specialized knowlege (Good for bottom line, bad for me right?)  VDI gives you that "One to Many", single touch point familar to TS while at the same time deploying into an environment that everybody knows including the vendors that support an organization's applications.  Obviously we are paying a premium now for that technology so maybe I'm still "cheaper"

appdetective wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 3:15 PM Link To This Comment

@caustic386, I'll keep it short as I have had my fill of posts today and need to work:-) It's been blogged about many times. Desktop Virtualization is all about the ability to isolate sessions and enable mobility/agility. You can't get this with static desktops. There is an upfront capital investment to do this that takes time to realize. If your business can move into countries, do move add changes, reduce BCP costs etc, this all helps offset much of that. It depends on what you do. In my case that is a lot of $$$$$$$$$$$ which makes up for some of the limitations which can be worked around. Over time as the mgmt tools mature the single image layer cake thing will likely happen to reduce OpEx. If that is your day one goal forget about it, although I will argue that there are OpEx reductions to be gained by centralization. So focus on the word agile and see how that fits with your business today. Ok no more posts today!

caustic386 wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 4:42 PM Link To This Comment

@appdetective - I think we're in complete agreement.  What I'm confused about, is why you feel that hypervisor-based VDI solution might be preferrable to something like Parallels?  There are certainly a few scenarios I can think of, but not many (certain driver issues, for instance).

Shawn Bass wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 5:58 PM Link To This Comment

You're all cracked.  Screw the hypervisor, screw the broker, screw the desktop personality and workspace management crap.  What am I doing?  I'm rewriting all of my apps for this new thing called the Web.  It's a beautiful thing.  All I have to do is rewrite all of my apps and boom!  Desktop Management problems solved.  While you guys are still struggling with how to manage your desktops I'll be sipping a fruity rum punch on the beach.  So it may take me until 2024 to get my apps all rewritten, who cares?  I'll have the last laugh... errr...

Shawn

Ormond Merino wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 6:15 PM Link To This Comment

Amen Shawn....It wont be soon enough when all these apps are simply browser objects.

Why have the burden of another user interface (VDI) when every end user device already has one?

Virtual Bridges wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Wed, Aug 12 2009 6:54 PM Link To This Comment

I respectfully contend that the “Madden Paradox” is fundamentally flawed because it is equating “VDI” with VMware and Citrix and not taking into account more complete offerings in the market such as VERDE from Virtual Bridges and IBM.

IBM gave a very compelling demonstration and presentation of the newly-announced VERDE 2.0 yesterday at CloudWorld in San Francisco.

If the world were flat then yes, the “paradox” is increasingly true. The gap between TS and VDI is increasingly negligible. But, the world is not flat, and there are vendors out there who offer truly innovative VDI, not just blatant extensions of the server consolidation paradigm with token efforts at incorporating innovations that address desktop-centric computing. Of the two, Citrix at least seems to understand the desktop better but they are the ones most caught in the teeth of the “Madden Paradox” – but that is another discussion.

So, if you come to see the world as round, you will see that Virtual Bridges’ VERDE has been designed from the ground up as a pure VDI product and addresses most of the issues that you have been insightfully pointing out are wrong with the flat-world comparison for TS and VDI.

When looking at The Paradox in the context of TS versus VDI, where VDI is defined as VERDE, then there are a number of distinguishing advantages of VDI over TS – and, I concede that TS still fills certain use cases better than VDI, but in meeting the needs of next generation desktop management, which is what VDI is all about, VDI has a number of distinct advantages:

1) VDI is superior for Public Clouds and MSP’s because of the better inherent segregation of users on the same servers. This is more cost efficient (HW, SW and operating cost), more energy efficient and more flexible for serving diverse populations efficiently.

2) VDI is superior in its ability to accommodate heterogeneous user environments - VDI allows one to seamlessly deploy a mix of XP, Windows 7 and Linux on the same infrastructure. This is important to not only future-proof your management and provisioning infrastructure, but also to make it more flexible in assigning the right desktop to the right job instead of a one-size-fits-all tariff on your desktop population.

3) VDI is superior in Application compatibility… (as referenced above). Not only are there applications that are not TS-aware or TS-friendly, but when dealing with sensitive legacy applications with finicky food chain dependencies, it is a lot easier/cheaper to isolate and provision these from a separate gold image than it is from a separate server as is required in TS.

4) VDI is superior in enabling disconnected use/local processing. Just because VMware and Citrix have no figured this out yet doesn't mean others haven't. VERDE has had support for disconnected use for several months. TS will never cross this barrier. VDI (VERDE) can do it today.

5) VDI is more flexible in accommodating distributed infrastructure whereas TS requires a centralized monolithic array of servers. This is currently unique to VB, and counter-intuitive to VDI, but the ability to have regional data centers and replicated branches that are centrally managed and dynamically provisioned as opposed as having to consolidate infrastructure into data centers gives VDI a practical advantage in situations where network reliability or bandwidth is an issue.

6) VDI servers are lower maintenance – VDI architectures are effective stateless when it comes to applications on servers whereas TS architecture needs to have applications installed on each server, and each server therefore requires more management and care. VDI involves running of desktop gold master images that are stored on disk. The applications are installed in the gold master images, not on the server.

7) VERDE VDI has built-in storage optimization. With its minimal persistent storage requirements this compares a lot closer to TS than Vmware for example. This isn’t really an advantage of VDI over TS, but it narrows the TS advantage

8) In the same vein, unlike the re-purposed server-oriented hypervisors like Xen and ESX, VERDE uses a more modern hypervisor approach so that densities of our desktop-oriented hypervisor are closer to TS densities than to the densities of antiquated server-oriented hypervisor densities.

So, in our view, VDI has several compelling and sustainable advantages over Terminal Services that will allow it to fulfill the high expectations that analysts and strategists have for it… but it will be a slow painful process if thought leaders continue to look to Vmware and Citrix for delivering on the promise.

Unfortunately, if you concede that there is a bigger world out there beyond just Vmware and Citrix, you will have to give up on your dream of the “The Madden Paradox” being your ticket to the Industry Lexicon Hall of Fame… ;-)

Martin Ingram wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Thu, Aug 13 2009 10:07 AM Link To This Comment

"My point is this paradox only applies if you’re using VDI because it’s easier for users to install their own apps and own their own environment."

OK, but is it just user installed apps? My suspicion is that if we had the capability for users to install applications on TS, then there would be other reasons that a seperate operating system per user still made sense. That is before I start even thinking about how we would manage applications that need to install kernel level components, services, etc. on TS.

There is more heretical comment from one of our guys here:

appsense.wordpress.com/.../ts-vdi-are-we-really-needing-vdi-yet-doesnt-ts-deliver-similar-stuff

Martin Ingram (AppSense)

Joe Shonk wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Thu, Aug 13 2009 10:58 AM Link To This Comment

How about the "VDI Complexity Syndrome"

As far as Parallels goes.  They are in weird position.   They can offer some of the benefits of VDI with some of the of TS.  Such that, I don't think it makes sense in most cases.  Either you have a strong case for TS or a strong case for VDI.

Parallels is still a server OS and not a Desktop OS.  Here you've just negated a use case for VDI in which you need a XP/Vista for application compatibility and vendor support.  So you need a real desktop OS, not one dressed up to look like one.

Not everything WILL run in Parallels.  Parallels has bandaids and shims programmed in to make some applications work.  What happens when the application is upgraded or the banaids fall off?  One prime example is XenApp 4.5 R04.  So when they get R04 finally working then we'll be on R06. ;)

XD provides USB redirection.  Parallels?  Nope, XD doesn't install on a Server OS.

Parallels needs TSCals instead of VECD License.  But for Parallels, you also need a DataCenter OS License per proc.  Microsoft is making sure they make their money regardless of which solution you have chosen.

Joe

Claudio Rodrigues wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Thu, Aug 13 2009 10:59 AM Link To This Comment

@Virtual Bridges. From what I remember VERDE is linux based meaning it does provide access to Windows applications but not running on a native Windows OS. That probably introduces compatibility issues like other similar solutions (not similar in all aspects but similar in allowing you to run Windows apps on non-Windows OSs) like CodeWeavers CrossOver products. Is that the case?

If it is I am almost certain that not all Windows apps will work what introduces a problem that does not exist when using a true Windows OS virtualized guest.

John Radcliffe wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Thu, Aug 13 2009 11:16 AM Link To This Comment

VirtualBridges - unless I'm missing something in the press release, this product is about supplying Linux desktops to users.  Great idea, but kind of leaves out the 90% of Windows apps that people actually use to conduct business.

Concerning the rest - What is meant by giving users the ability to 'personalize their desktop'?  Are we talking about the base things like screensavers, background, etc? (don't think so) or giving the users the ability to install apps that they feel they need along with customizing the desktop settings?  It's a great concept to let users decide what they want to install, but it's an administrator's nightmare from support, security and licensing.

Now, installing pre-approved apps on-demand - that's doable for both TS and VDI.  Not going to go into the TS vs VDI thing here other than to say that there are reasons to use both and often they are used in combination (it's not an either/or).  It's not an easy to support these technologies, but giving users free reign on their desktops VDI or TS is just bad for everyone involved, including the users.

Brian Madden wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Thu, Aug 13 2009 12:10 PM Link To This Comment

Yes, remember this is just about server-based computing (since if it wasn't then you wouldn't be talking about Terminal Server). So this is why I defined VDI as server-hosted desktops, which is where I think this paradox comes into play.

If you're talking about general virtualization of desktops, then that's a whole different story and the paradox doesn't apply because it's not just about TS.

Virtual Bridges wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Thu, Aug 13 2009 12:22 PM Link To This Comment

Thanks for the comments/questions, John and Claudio...

Claudio, first… VERDE is not similar to WINE… We are and always have been a true virtualization solution. VERDE functions as other VDI solutions in that it runs a full copy of Windows in a virtual machine, it just so happens that instead of the VM running on a Windows server or ESX (Linux roots); it runs on Linux servers. There are many advantages to this approach in addition to cost, security and stability – less memory required of the VM’s, more flexible file system choices (any shared storage, not just SAN), better hardware densities, etc… . So, the result is that we have the same 100% fidelity with desktop applications compared to other VDI solutions, but we do it more efficiently with Linux servers.

To John…  VERDE runs BOTH Windows and Linux guests from the same infrastructure. In other words, on the same server, or cluster of servers, you can have any combination of Windows virtual desktops and Linux virtual desktops. We will be introducing Windows 7 very shortly as well.

The press release is geared towards IBM's message of an “MS-free environment” but if you read closely you will see that we got the Windows capability in there as well ;-) We believe that the market for this initially is WINDOWS and if an organization wants to further lower costs by segmenting their population to optimize licensing costs for certain use cases, we support that. Flexibility is important. With the rapid rise of the Cloud and more SaaS and network-delivered applications, the robustness of the Windows desktop environment may not be as appealing in the future. Why lock into an investment in Vmware or Citrix for large amounts of money that locks you into the Windows thick desktop model. Having options is a good thing – even if only negotiating lower Windows fees by demonstrating a credible threat of force… and believe me, we see this happening more and more.

On a related point -- There was a story in Network World this week about MS killing VDI with the high costs of VECD. I have two comments…

First – the case for Linux desktops:  You can’t blame the steak for your heart attack. You have a choice as to what you put in your mouth. Mixing metaphors, don’t throw the baby out with the bathwater – just because Windows VECD is considered expensive, doesn’t mean that VDI is going to fail… at some point organizations might realize that they can change their diet and get by just fine, and a lot healthier, economically. Linux is increasingly an option in these frugal times.

Second - In defense of Microsoft: Why pick on Microsoft? Looking at VDI, there are two parts to the equation – the desktop and the infrastructure. If you were to adopt VERDE instead of Vmware or Citrix, you would save enough money on both SW and HW acquisitions costs that even with carrying the extra cost of a VECD, you would still be saving considerable money. So, the only reason to blame Microsoft for killing VDI is if you accept the premise that you have no option but to pay $250+ per user for the SW infrastructure. With VERDE, the SW AND Hardware costs combined are often less than $200 per user.

So, blaming Microsoft for the cost of VDI is picking on an easy target, but by Network World exonerating Vmware and Citrix in the high cost of VDI, they were either uninformed or being disingenuous. At least Microsoft is delivering something the users want.

Rob Nicholson wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Thu, Aug 13 2009 4:41 PM Link To This Comment

We're a XenApp based small business. We use it as our destkop. It's allowed us to set-up a London and USA office for incredibly small costs. The TCO over ten years has been probably well over £1m. But we considering moving away from it as our desktop mainly because of multimedia concerns. That's the only reason. We can get around the other (few) TS issues. VDI isn't even on our radar as it suffers (today) from the sam issue - the mutlimedia experience isn't good enough. I'm confused why anyone would consider desktop VDI over TS if they can put up with lower mmedia experience. The TCO doesn't stack up.

Shawn Bass wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Thu, Aug 13 2009 10:37 PM Link To This Comment

@VirtualBridges - I'm all over the notion of desktop in the cloud, believe me I am.  However, there's this little thing called latency that sort of ruins it for me.  I mean sure if you running all apps where the infrastructure is completely hosted in the cloud provider and/or your apps are all fairly stateless and WAN friendly, then I'm all game.  But as long as the notion of centralized computing is needed to deal with crappy apps, then you're still going to need the execution tier to be near the data tier.  Futhermore, if none of my apps have problems and they all remote over the WAN, why exactly am I doing hosted desktops? ;)  Last comment is tongue in cheek because I do realize the benefits aside from the app/OS deployment benefits.

Shawn

Simon Bramfitt wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Fri, Aug 14 2009 3:17 AM Link To This Comment

Brian

You might be right in saying "this paradox only applies if you’re using VDI [VHD] because it’s easier for users to install their own apps and own their own environment".  

But is anyone actually using or considering VHD in this way?  

Simon Bramfitt wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Fri, Aug 14 2009 3:53 AM Link To This Comment

@VirtualBridges

I'm sorry Jim you can't start off making statements like " where VDI is defined as VERDE, then there are a number of distinguishing advantages of VDI over TS..." and then list the advantages that any VHD solution has have over TS as if these advantages were unique to VERDE.  

(Not unless you work in marketing anyway).

If VERDE has genuine advantages over other platforms, let's here them.  And that means providing a little meat beyond "VERDE uses a more modern hypervisor approach so that densities of our desktop-oriented hypervisor are closer to TS densities than to the densities of antiquated server-oriented hypervisor densities."  

Virtual Bridges wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Fri, Aug 14 2009 11:54 AM Link To This Comment

Shawn,

I think a TS mentality focuses too much on the apps and not enough on the user environment, i.e. the desktop. With a desktop model, a common set or sets of apps can be defined and managed at a lower cost than on an app-by-app basis, multiplied user-by-user, creating a matrix of complexity that increases exponentially with scale. Dealing with this level of complexity has been an inhibitor to broader adoption for an organization whose prime objective is providing users with the tools they need at the lowest possible cost. Desktops are a convenient organizing model for aggregating apps compared to a sprawl of app servers with their associated user relationships. So, as an organizing principle for the majority of users, “desktops” work.

In practice, today, however, desktops suffer from a lack of efficient central provisioning and management… and from their original sin of being ‘stateful’…Diversity in humanity is a wonderful thing; in desktop computers, not so much. If organizations could roll back the clock and make one improvement in the creation of the desktop, it would be to make the operating system and the applications stateless – pristine gold images crafted and tested by professionals would synchronize with legions of desktops automatically, being updated behind the scenes without a user ever knowing how, when or why this is happening; and the installation of applications would be the exclusive purview of the administrator – thus bifurcating the desktop computing world into two genres; the Personal Computer and the Work Computer. (BYOC anyone?)

Users should come to expect their systems to be state-of-the-art within the boundaries of corporate policy. Many users reading this might scoff at the marginal time and effort they are being saved by this, but, think of the vast majority of people who struggle with the simplest of maintenance tasks… Thus virtualization and automation offer huge cost savings when compared to today’s high cost of desktop ownership.

Regarding your concerns about latency… Virtual Bridges has a couple of unique capabilities based on our SMART protocol. SMART stands for Self-Managing Auto-Replicating Technology.

The first advantage to VERDE is that it manages a client-side hypervisor from the same VERDE infrastructure that supports traditional VDI users – so the benefits of gold master provisioning can be extended to the local machine. By efficiently replicating the gold image from the server onto the local machine and storing that as a write-protected replica of the gold image, all of the advantages of uniform imaging and stateless management are extended to the local processing model. So, no Cloud/WAN latency. The gold master runs on the client-side hypervisor and is automatically synchronized with the gold master on the network -- the same gold master(s) that the traditional VDI, server-based user would run off of.

The second way that SMART solves the latency problem is through a hierarchical image replication model. Instead of the scenario above where the SMART protocol is updating the local machine, where the local machine is a laptop, notebook, workstation, etc… the SMART protocol would update the replica of the gold master on a branch server, appliance or local cluster of servers. This allows the gold images to be centrally managed but automatically distributed to local proximity servers where the replicated image now serves as the gold master for VDI and disconnected users on the local LAN segment. Again, no Cloud/WAN latency.

Thanks for your question. Feel free to request an eval at www.vbridges.com

Virtual Bridges wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Fri, Aug 14 2009 1:53 PM Link To This Comment

Hi Simon,

I concede that my argument mixed together some generic VDI advantages with VERDE-specific VDI advantages in responding to the TS versus VDI discussion. Let me separate them out for you.

Points 1 (Segregation of users), 2 (heterogeneous environments), 3 (Application compatibility) and 6 (lower maintenance) are pretty generic. I will not overstate the importance of supporting Linux as this is still a niche market. The only claim I will make with respect to these points is that VERDE is easier-to-use, less complicated in design and available at much lower cost, thereby strengthening a VERDE comparison to TS versus a Citrix/Vmware comparison to TS.

Point 7 is more of a reference to Vmware. I am not sure how the recent announcement by Citrix provisioning from a gold master works (as opposed to requiring persistent, stateful VMs as Vmware does) but it sounds a lot like how we have been doing it for the past five years. And, I did concede that Citrix is doing a better job of trying to incorporate some desktop-centric design improvements – again, things we have been doing for some time that are well integrated and seamlessly managed in our product.

Points 4, 5, and 8 are unique to Virtual Bridges’ VERDE.

4 – Disconnected use/local processing. VERDE provides an integrated client-side hypervisor mechanism that is managed from the same VDI infrastructure as the server-based sessions. I have heard rumors of Vmware and Citrix working on their own, but ours is in deployment.

5 – Replicated branch infrastructure. With our SMART protocol we are able to automatically update/replicate gold images running on branch servers and appliances so that organizations can centrally managed their gold images but can have these images accessed locally from branch servers without having to incur latency penalties over the WAN. This combines all the benefits of central management with the performance and convenience of LAN-based delivery without having to manage the pushing or synchronizing of the images running in the branches, regional data centers, remote clusters, or in the case of an MSP, at their client’s site.

8 – Since VMware does not allow comparative independent benchmarking, we are going by customer experience. For example, where a customer may get 40 Vmware sessions on a server, with VERDE they are running 80 or more on the same server. If anyone would anyone do their own comparisons, we would be happy to work with them.

In summary, on the surface there are some generic similarities in several of the points I mentioned but when users actually compare the ease of deployment, the level of maturity and the cost differences, together with the unique capabilities, we believe VERDE is in a class by itself.

Simon Bramfitt wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Fri, Aug 14 2009 7:08 PM Link To This Comment

@VirtualBridges

Thanks Jim,

So it sounds like VERDE has many of the capabilities seen in other products but is the only product available today that presents them all as a single product.  It looks like you might have a fight on your hands with Citrix though, this sounds very much like the direction that Citrix are taking with XenDesktop when coupled with XenClient. With the significant differentiator that VERDE is shipping and Citrix is still in the development stages.

it certainly warrants closer inspection.

Simon

Shawn Bass wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Fri, Aug 14 2009 7:48 PM Link To This Comment

@Jim - Virtual Bridges -

It's ALWAYS been about the apps, it's currently about the apps, and it probably will be about the apps at least for the next 1-15 years until the SaaS/cloud model is omnipresent.

Let me qualify this by saying I do 99% of my consulting in the enterprise space where you're dealing with 10k+ desktops, 1k+ applications, and hundreds or thousands of servers.  I just can't see any way that these types of customers could adopt a hosted desktop solution like you're advocating without a complete data center relocation into the "cloud" provider.  Maybe once all the apps are SaaS this is doable, but now not a chance.

But then again perhaps your market focus is SMB in which case I suppose it's completely plausible (assuming you can still provide virtualized servers in your data center for those applications that do require a client/server approach).

But keep in mind, not everyone out there just runs MS Office, Notepad and calculator.  A desktop (and the APPS specifically) are much richer from an infrastructure perspective. Don't oversimplify that with some marketing bullshit.

Shawn

Virtual Bridges wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Sat, Aug 15 2009 2:48 PM Link To This Comment

I think we have a bit of a disconnect here… we are talking about virtualizing desktops in the private cloud as well as the public cloud. Some of your concerns sound like you think we are talking about moving the entire enterprise infrastructure into a 3rd party hosted public cloud model… No, we are not advocating that for the large enterprises, at least not yet… for large enterprises, we recommend implementing VDI in a private cloud, i.e. the existing data center. For SMB, and for organizations that already rely on MSP’s and outsourced desktop firms, yes, this is a great public cloud solution that will lower the cost of providing these services for both parties.  

In a private cloud implementation, we are not talking about changing anything about how the data center runs wrt back office servers or current data center models (i.e. Vmware server consolidation). We are talking about an added enterprise data center capability to accommodate VDI, where VDI has very different characteristics as compared to server consolidation.

So the client-server comment puzzles me (unless you were assuming we were talking about moving this to a public cloud). We run the client piece of the client-server application in the virtual desktop, just as it runs on a physical desktop today. The only difference is that the “client” piece of the client-server application would be in a gold master image, together with whatever other apps the user required, and would be managed centrally as a gold master image instead of managed as a stateful endpoint. The server part of the client-server application would be unaffected and run in whatever way it runs today. That is outside our scope.

Our, or any, VDI approach certainly doesn’t preclude large numbers of apps or large numbers of users. We are very enterprise focused – I have been in enterprise infrastructure my whole career. VDI is simply a new way of defining the relationship between administrators and the desktops they administer. Instead of managing multiple stateful endpoints, organizations are able to define a more streamlined set of gold images and associate users and groups to these gold masters. The value of the virtualization is in at least three areas; abstracting the idiosyncrasies of the endpoint hardware out of the equation by presenting the desktop image with a consistent view of the hardware (virtual machine); second, locking down the desktop image so that users and malware cannot effect persistent corruption/changes to the image – if an error or corruption occurs, users simply restart the session and the write-protected pristine gold image is re-instantiated in seconds. The third benefit is that administrators, and by extension the organization, have control over what runs in that image so users have a consistent environment with all the necessary apps, patches, versions, etc…in a known working state (i.e. tested and certified to have no conflicts), which can be universally updated as easily as updating the gold image(s). Of course, when Windows 7 comes out, updating your entire organization will be as easy as updating the respective gold image(s). Another benefit of this is that if there ever is a need for a roll-back, for whatever reason, it is just as simple. There is no push involved with this type of VDI model. When the gold master is updated, users automatically inherit the new image when they start their next session. This is not “marketing bullshit”, this is how proper VDI works.

BTW – I think the TS niche is safe from extinction. We may agree that XenDesktop is making some interesting noise, but, as I said earlier, I think Citrix is caught in the horns of “The Madden Paradox” more than most. Practically speaking, it is hard to put the stud of the stable out to pasture when it still has plenty of giddy-up.

appdetective wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Sun, Aug 16 2009 10:50 PM Link To This Comment

@Virtual bridges. What is your display protocol? IBM is unable to do anything in this area. Look at the Cluster F with Net2Display. You're extending the user presentation further away from the physical so protocol is everything irrespective of backend placement. If your answer is a distributed storage architecture, you are are complicated and I guess $$$$$ out of the gate. How is that different from Unidesk or Atlantis?

@caustic386. Parallels can't patch at the same rate as MS. I not going to trust that architecture and there is a zero day security exploit... Joe Shonk above mentions a bunch of app compat and os compat stuff as well. In addition I will add MS is making TS better with more isolation, so this model is not something they are investment in, so neither should anybody else in their right mind. If you do, great , good luck.....

@Claudio Rodrigues If your barrier to adoption to VDI today is cost don't do it. I've had similar conversations with Citrix, and I think they would agree that investing in a centralized infrastructure is a forward looking view to drive towards efficient management, vs. the short term view that it's capital upfront and panic which is what most people do. Yeah there are limitations like multimedia but they will get better over time. The Agility is what it's all about. The costs will come down over time. So it's a choice, sit and wait, debate and watch while others get their organization reshaped. Citrix gets that!

Shawn Bass wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Mon, Aug 17 2009 11:12 AM Link To This Comment

@Jim (Virtual Bridges) -

Perhaps I did misunderstand because I took it as a public cloud offering.  If it is truly a technology that anyone can purchase and implement in their own data center, then my concerns around latency, etc. are not there.  However, I will start asking other questions such as:

1) What is your display remoting protocol (already asked by AppDetective)

2) Where did you guys come up with your numbers?  (250 concurrent sessions per server/blade - what kind of hardware were you running this on???)

Shawn

Claudio Rodrigues wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Mon, Aug 17 2009 12:46 PM Link To This Comment

@appdetective I am not saying it is my barrier. I am just saying that as any other piece of technology, when VDI is considered, cost is a factor for sure. When the economy is doing great the 'savings down the road' argument is plausible. But when things are pretty dry, it gets very hard to justify anything that has a big upfront cost, no matter how much you show on paper the savings over the years.

Virtual Bridges wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Mon, Aug 17 2009 9:21 PM Link To This Comment

@appdetective and @shawnbass

Regarding display protocol:

Our display protocol is an open source implementation combining many technologies.  We use a multi-channel approach to provide access to local resources as well as display.  On the display channel, we use a derivative of the RFB (Remote Framebuffer) protocol that also does JPEG compression to reduce bandwidth requirements.  The display channel does not connect into the virtual machine - instead it displays the VM's virtual video device so it does not require any drivers inside the guest and supports any guest type (including Linux).  On the other channels, we provide a protocol we developed years ago called CDA (Client Device Adapter) which enables things like audio/multimedia playback, universal printing, and file sharing.  We deliver this over 128-bit SSL utilizing the Blowfish cipher, which is very fast and very secure.  Our client software is available in source form, in Linux binary form, and in Windows binary form.  It has also been ported to Solaris (including Sun Ray) and to various specialized thin client platforms over the years.  It is entirely software based and runs on virtually any thick or thin client device, with add-ons available for the likes of Wyse and DevonIT through our IBM relationship.

Our protocol is capable of delivering rich multimedia playback over broadband/WAN connections, and can even deliver a decent business application experience over intercontinental latencies with minimal degradation.  Of course to deliver rich multimedia across these intercontinental, high latency distances we support the centrally-managed replicated remote site model (see below), but this is not strictly necessary for most business applications, so you can keep implementation costs low even with a global user base.

Regarding our replicated remote site architecture:

Our solution is not a complicated, expensive distributed storage architecture. It involves what might be called off-line, or background processing to synchronize a small set of files/images not requiring performance or real-time processing…. Through our SMART protocol we are able to automatically synchronize images maintained at a central location with multiple remote locations thus giving the organization the benefits of central management and the users the benefit of local performance.

Difference with Unidesk and Atlantis:

Unidesk and Atlantis are eco-system products. They require the connection broker and infrastructure of Vmware or Citrix. The elegance and simplicity of being a fully-integrated solution is our primary technical differentiation… For the functionality we are comparing, VERDE does the same thing in one stack that VMWare/Citrix plus Unidesk/Atlantis might do.  From a more fundamental perspective, I think cost is the ultimate differentiator. We provide the whole integrated stack for a 1000 users and above at $50 per seat – no add-ons, no up-sells, no premium editions, no recurring subscription fees; It is a paid-up license that includes everything we have - $50 per seat. So, not only do we deliver superior operational efficiency and cost saving value, but we deliver it at an exceptionally low acquisition cost as well.

@seanbass… I am not sure what you are referring to when you ask about “250 users per server”. This is the only line in the thread I could find with “250” in it:

“So, the only reason to blame Microsoft for killing VDI is if you accept the premise that you have no option but to pay $250+ per user for the SW infrastructure. With VERDE, the SW AND Hardware costs combined are often less than $200 per user.”

I was not talking about users per server, I was talking about a published cost per seat of our competitors. And, so as to avoid confusion - the $200 number is assuming $50 for the SW and roughly a $150 slice of server HW per user.

appdetective wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Mon, Aug 17 2009 9:52 PM Link To This Comment

@virtual bridges. I would love to see your product reviewed. Good to see innovation, so I'll keep an open mind. Good luck.

Azam wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Fri, Aug 21 2009 1:34 PM Link To This Comment

I am not sure why Citrix does not emphasize more about the streaming technology it has built in XenApp. The biggest advantage of streaming applications to client devices is to take some load off from data centers and off-line access. The client devices these days are powerful enough to handle some load themselves. What streaming technology needs to do is build a feature which will analyze a client device to ensure that it is ready to download a streaming package if it has enough RAM, space etc., and run it locally otherwise run it from the server. Offline access in itslef is so powerful that I think it can help Citrix make great inroads to many new markets when it comes to mobility and offline resources availability. Another smart access or use of technology would be if the internet link is slow or bad switch to offline usage instead of online usage. I love the fact that you don't need to stay connected to a server in order to use an application or a resource available on a server.

When it comes to VDI solutions a feature like streaming will come in handy as well. It is a great deal for CIOs and IT managers to ensure that their users can access resources 24/7 365 days no matter where they are. If a machine can handle then stream the virtual desktop to the client device or fit it on a USB for mobility purposes or bad/no internet links. I think MocaFive has done a great job when it comes to managing or delivering virtual desktops to client PCs either via streaming or USBs.

appdetective wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Mon, Aug 24 2009 9:12 PM Link To This Comment

@azam the only think I wonder about the ability to dynamically decide where to execute is the user support calls. I'm not sure they would understand the difference and complain at the different experience. Perhaps an option to get the user to pick would a better to force the training issue.

Shawn Bass wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Fri, Sep 4 2009 11:11 PM Link To This Comment

@Virtual Bridges -

I got the 250 users per server blade straight from your Verde Cluster overview PDF over here:  vbridges.com/.../VERDEClusterOverview.pdf

You say on page 4:

-Host up to 250 concurrent sessions per server or Blade

-Manage up to 10,000 (ten thousand) servers or Blades in a single cluster

-Serve up to 1,000,000 (one million) concurrent users per cluster

Of course you do throw a clause out there that says "Please note that these are absolute, maximum limits. Practical limits may be lower due to processing power, storage capacity, bandwidth, application profile, and use case."

I guess that's kind of like saying "I could put 3,000 concurrent desktops on a single ESX server if only I wasn't limited by CPU, Memory, Disk and Network"

But, I'll concede that I didn't understand your product before, so perhaps I'm not understanding what this 250 concurrent users to blade business is all about either.

Shawn

Michael Rueefli wrote re: Introducing “Madden’s Paradox”: the gotcha of the VDI versus TS debate
on Sat, Jan 30 2010 7:56 AM Link To This Comment

I believe questioning the end user experience if you talk about VDI today is one of the most important things. If a company's users are used to have mostly actual HW on their endpoint side, you won't make them happy if you force them to use a virtual device instead. Of course, you can tune your VDI environment accordingly, but that doesn't save money at the end. Shared, perforfmant storage, as well as a bunch of performant servers are required. So it ends up there, if your users claim about performance, and you have to decide to ignore that because of having a central managed endpoint, or to give them back a physical endpoint. I know, I know, the vendors tell you that you can also centralize physical desktops in the datacenter, but come on... who want's to give away expensive datacenter space to a few desktop hardware pieces?

Managing the user app and profile environment is not that different from having a traditional desktop environment, but it also ends up at the performance and end user experience.

I don't think finally that you can compare TS and VDI solutions. Each of them has it's own benefits and issues you'll run into as you have to design an infrastructure for thousands of users.

I agree, that the brokers have to grow up a bit, there are too many issues today using them, but of course at least you have to use them, otherwise you'll loose a lot more of time managing the delivery of virtual desktops to your users.

(Note: You must be logged in to post a comment.)

If you log in and nothing happens, delete your cookies from BrianMadden.com and try again. Sorry about that, but we had to make a one-time change to the cookie path when we migrated web servers.