Will VDI replace Terminal Server-based SBC?

Last week Computer Reseller News published an article by Mario Morejon titled, "Virtualization: The Emerging Desktop." Quoting from the article: "Virtualized desktop technology holds more promise than server-based computing because users are able to experience fully integrated desktop environments.

Last week Computer Reseller News published an article by Mario Morejon titled, "Virtualization: The Emerging Desktop." Quoting from the article: "Virtualized desktop technology holds more promise than server-based computing because users are able to experience fully integrated desktop environments. Personalization is relatively easy to set up. This is why Citrix is now in the game. Server-based computing, including Microsoft Terminal Services, is a dying technology."

My initial reaction was "No. No. Wrong. BS. Nope. No way. Nuh-uh."

Besides, the article was so completely full of inaccuracies and the author obviously has no clue about this space, so I laughed it off and ignored it. But a few days later I realized I was still thinking about it.

And now I ask myself, "Is this possible? Could VDI ultimately replace Terminal Server?"

This blog entry is two parts--both in reaction to Morejon's article. In the first part I'll explore whether VDI could replace Terminal Server. In the second part, I'll point out a few of the big errors in the CRN article. My fear is that people not "in the know" (i.e. our managers) will read it and get all sorts of mixed-up notions about VDI and Terminal Server.

Part 1. Could VDI replace Terminal Server?

For those readers not familiar with my views, let me first state clearly that my bias is towards Terminal Server. Sure, I'm independent and a consultant and always looking out for the best for the people I coach, but let's be honest: I've been working with Terminal Server for ten yers. This "VDI thing" kind of snuck up on me, and at first I didn't really think it had legs at all. But as you might remember from my most controversial article ever, I'm definitely seeing certain scenarios where VDI could be useful. So my "VDI Love" index is increasing.

(Of course "VDI" means a lot of different things, depending on who you ask. For the purposes of this article, I'm using the term "VDI" to mean Windows XP or Vista instances running in a data center, with users connecting to them remotely via a protocol like RDP.)

Could VDI replace Terminal Server-based solutions for SBC? As I see it now, there have traditionally been two major reasons to use TS instead of VDI:

  • Given the same hardware, you can typically fit more TS user sessions than VDI Windows XP VMs.
  • Since VDI means running a copy of Windows for each user, you have to figure out how to manage all of those Windows instances. With TS you only manage a single OS instance for all the users on a server.

I've written about these two reasons in the past and how I think they're evolving, so it's not really worth re-hashing all of the specifics here. But some quick notes for those who don't want the full dissertation:

First, recent changes in hardware have allowed VDI to "narrow the gap" in terms of number of users per server. Now that most servers are dual and quad core (and moving higher), VMware/Xen/Hyper-V can / might scale up on those boxes as good as (or better even?) than Terminal Server.

Second, if you have some method for managing disk images for your VDI users, the whole "managing all those Windows XP images" isn't too big of a problem. There are lots of ways to manage these disk images: Citrix Provisioning Server (Ardence), disk image snapshotting, etc.

Some people say "well sure, you can manage the images as one if they are stateless, but stateless won't work in our environment." I agree 100%. If stateless doesn't work, then VDI image management will be a nightmare. But the point that I'm making is not comparing VDI to the traditional method. I'm comparing VDI to Terminal Server, and all Terminal Server sessions are (by definition) stateless since all users on the server share the same disk image.

Assuming those two statements are true, the advantages that TS has over VDI are diminished. In that case you could go VDI and get the benefits of Terminal Server and also get the benefit of not having to worry about Terminal Server app compatibility. Plus you could probably hook into something like Citrix Provisioning Server or VMware ACE and "stream" the actual VDI virtual machines to a client device where they could be used offline. You could also move them between servers.

So thinking like that, yeah, I do wonder whether the future of Terminal Server is limited, and whether a single user instance OS (Vista or XP)-based VDI solution might be more practical at some point in the future. (I don't think we're there yet. Maybe in a few years?)

Part 2. That article in CRN was the worst piece of garbage I've ever read

I really don't want to get into a line-by-line rebuttal of this article. And as you just read in Part 1, the ironic thing is that I agree with Morejon's basic idea. But the examples in the article are ignorant, dangerous, and just plain wrong. A few of the major errors:

In the first few sentences, Morejon repeatedly misspelled "Xen" as "Zen." This indicates to me that he doesn't have too much experience in this space, and makes me question his authority for this article.

He goes on: The full desktop experience has never been fully captured within server-based computing. Virtualized desktops do a better job at making users feel that they are in a more normal environment and users have more control over work environments. By contrast, server-based computing provides a highly controlled and locked-down experience. Users only get access to what administrators deem necessary.

This is a classic mistake. He is comparing "virtualized desktops" to "server-based computing." These are not opposites! Virtualized desktops (in the VDI way he describes) is server-based computing! It's just that VDI is server-based computing connecting to a single user OS (Windows XP or Vista), whereas the more traditional Terminal Server-based server-based computing connects back to a multiuser OS (Terminal Server.)

Of course he also talks about "SBC" (i.e. Terminal Server) as being more controlled and locked down, and that users only get what admins deem necessary." I don't think he realizes that TS admins do this on purpose. I really hope that admins also apply this approach to VDI. (But I get the point that with VDI you can be more loose since if a user breaks something, he only impacts himself.)

With virtualization, users do not need to have costly, full-power PC systems on their desks. Since the working experience is coming off a virtual environment, users can have inexpensive thin clients and still have the same experience. Thin client devices even provide more access to remote software installations.

I agree, except he's using this as an advantage of "VDI-based SBC." In reality this is an advantage of all SBC solutions, whether single-user VDI or multi-user Terminal Server, so it doesn't really support his case against TS.

"Companies are finding out that virtual desktops are a logical solution for remote locations," said Tom Cecchini, president of Async Associates, Wixom, Mich. He added that technologies like video streaming, Reader Rabbit (in the educational community) and CAD systems have never been able to work on server-based computing. Thin client is changing the old perception by allowing users to have that power at remote sites."

Again, these problems apply to VDI and Terminal Server-based server-based computing solutions. Video streaming is a *** over SBC, regardless of whether your back-end is a single user OS or Terminal Server.

After trying an original Windows 2003, the installation failed. Engineers tried using virtual floppies on the remote PC manager running the VI client software. However, the virtual floppy drives did not activate and driver files could not be installed. Engineers do not recommend installing OSes into a virtualized environment. It is easier to use an OS image.

Morejon is basically saying "We couldn't figure out how to boot a Windows ISO to install Windows into a VM, so instead we recommend you install Windows onto physical hardware and then do a physical-to-virtual migration." Seriously? This is the team you want to take advice from? The "P2V is the preferred installation method" geniuses at this magazine?

This technology has come a long way since the days of Citrix's ICA

This scares me. Their demo environment used RDP! I feel that this statement is like a "dig" to Citrix, saying that RDP is much better since the "days" of ICA. Which "days" is he talking about? The days of 1998? Because if so, he's right! RDP of 2007 is better than ICA of 1998. (Just like a 2007 Mazda is faster than a 1982 BMW.) Apparently they don't know that ICA is still around.

But that's not the worst part of that statement. If he likes RDP so much better than ICA, fine. But he's using it to prove the point that VDI is better than TS. Except of course you can still use RDP to connect to TS.

Companies should see less traffic on corporate networks when using virtual thin client environments.

Oh man! It's going to take years to undo the damage of this inaccuracy. This statement is straight from the Citrix-angular-flying-briefcase-man of the 1990s. When this dude gets his reader rabbit and video streaming and CAD running via RDP, I can guarantee he's not going to see less traffic on his corporate network!

You get the idea.

The bottom line is I do really think the future of SBC could be more around single-user OS instances. But I sure don't think that because of anything Morejon wrote in that article.

By the way, this article is full of HP and Neoware (now owned by HP) references, including kind of weird references as to what model of HP laptop the engineers used to connect to the management consoles. I sure hope HP didn't pay for this crap!

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

Sorry, just had to go there. LOL


Good article Brian. I've not read the CRN article, but given your quotes from it, I doubt I ever will.

With regard to part 1: from my perspective what we are seeing is actually a convergence. If you take the view that the hypervisor is the new OS, with the virtualized Windows instance as the "application" or session, then TS and VDI are not so far apart. The choice becomes a technical one: how much are you willing to pay in terms of scalability and management for improved compatibility and isolation. That is why I'm in agreement with the view you've expressed in the past that the blended approach, where the administrator can pick and choose, is the correct one (and that is the approach we are pursuing).

I do have a question - you write:

... recent changes in hardware have allowed VDI to "narrow the gap" in terms of number of users per server. Now that most servers are dual and quad core (and moving higher), VMware/Xen/Hyper-V can / might scale up on those boxes as good as (or better even?) than Terminal Server

Obviously hardware improvements will enable putting more VMs on each server, but the same is true for TS sessions as well. How will hardware improvements enable VDI to achieve greater scalability than TS?


That was the most painful article I have ever read (ok, maybe not the MOST painful but close).  The fact that CRN even allowed that article to be published with all the incorrect statements is horrible.

The stuff about Reader Rabbit now being able to work better over VDI is the misconception that orginally got companies like IBM all jazzed up a couple of years ago about this space.  Unbelievable.

I agree there is a space for VDI.  The thing I don't understand is how it would replace SBC as we all know both technologies today.  I don't need a desktop and in reality, it's not like a task worker wakes up in the morning thinking, "Well, I've got to get that presentation done in Windows XP".  Instead they think, "Well, I've got to get that presentation done in PowerPoint.".  I think the reality is that virtualized applications will win over anything else as the extension over published applications.

Microsoft does $800 million on just Sharepoint sales alone.  This makes it the fastest growing server prodcut ever for Microsoft since its inception.  Companies are using that as their "landing page" in essence replacing the desktop and launching applications from there.  I'm not saying that Sharepoint is the end all, be all. . .just questioning why a desktop is so important which in essence is what VDI is giving.  Or, do people look at it because it is really "virtualized".  I think the latter and which is why virtualized apps will win out.



Thanks for another excellent and insightful post.  It's great to see someone who has so much TS/PS history and yet is not afraid to question (and re-question) the future of SBC.

Regarding your question "Could VDI replace TS?", I think the answer lies in what TS is being used for.  In the case of TS-based desktops, I think definitely and have explained the reasons why in the maiden posting on the Desktone blog @ www.desktopsasaservice.com.  The primary reason has to do with the balance between user experience and hardware resource sharing.  However, I do make it very clear that I think TS will continue to be successful as an application deployment solution for rich clients.  App deployment to rich clients accounts for the overwhelming majority of TS deployments.  TS has never really been strong as a desktop replacement solution for all the reasons you know so well.  The primary one here, again, is user experience and the fact that users need to work in an unfamiliar Windows Server environment as opposed to the Windows Client environment they've come to know and love.



VDI could replace TS in environments with a limited number of apps, but as the number applications grow a complete VDI environment would be difficult to maintain. I mean, think about it. Say you've got an HR image, an Accounting image, a Call Center image, etc, etc... then some employee in HR needs an application from the Accounting image. What are you going to do? Add that Accounting app to the HR image? Maybe, but can you allow the rest of HR to see that accounting app? Even if you could, is that accounting app compatible with all the other HR apps? Do you have the time and resources to test it first? Or do you just create a another image for that "one off" user? Maybe, but what happens if these types of "one off" scenarios occur again and again. Do you have enough VDI storage space for an increasing number of "one off" VDI images. Then what about that Call Center employee who is the only one in the company who needs a specific app? All the same questions need to be answered again. It comes down to how can you manage the "ad hoc" scenarios that occur in IT when users share the same source VDI image? If these "ad hoc" occurences are few and far between, then sure VDI could replace TS. But in larger environments these occurences would drive the admin of the VDI environment crazy. With TS and Citrix you just publish the app to the user with the "ad hoc" request. And with app streaming, the user can do something with one specific app that VDI can't do... take that one app offline. If you stream down an entire VDI image, they take all those apps offline, which likely isn't something you want them to do. But, VDI could have a significant roll in large envirnoments, but I think Citrix has got it right in that VDI should be used to deliver a desktop environment, but TS should be used to deliver the majority of the apps. If you have an app that isn't TS compatible, then you have the option of putting it on the VDI images. This would be the best of both worlds. 


...........you stream the applications based upon the users requirements to the same VDI Image!




I have to agree with you...this is where you would do mix of both, or perhaps do VDI with something like AppStream, but really I can't seem to figure out how the "VDI image management will be a nightmare."  I see the VDI images just as if they were a typical workstation, only virtualized.  I could see how VDI how and will replace any type on scenario where you have users use a thin client into a fat client for COOP purposes (I support a network that does this), but really I see places where both a terminal server and VDI are both needed, and that neither one will fade away. 



Citrix License - $500, TS License  - $60, High powered 2 CPU, 4GB RAM server 10K serving 50 users.

I am not sure what the VDI costs would be - I am asking for estimates. 


This is true on several levels with provisos:

  • 32-bit TS installations don't scale well past 50 users due to kernel/user memory limitations
  • Multi-core. multi-socket servers using virtualization will attain better overall CPU utilization than a single OS instance of W2K3. Scaling with virtualization is much more linear as Windows server code doesn't have linear performance gains with additional cores/sockets
  • A bunch of single vCPU Windows XP/Vista VMs will provide better overall CPU utilization

Countering the above, when virtualization of TS/Citrix servers gets better (I'm hoping that NPT technology from AMD Quad cores and VMware ESX 3.5 support for it will provide this) then the advantage of using VDI VM instances vs. 32-bit TS/CPS VM instances will be nullified.

Also, running TS/CPS 64-bit OSes natively makes better use of hardware and scales much better.

I think that ultimately neither SBC as it exists now nor VDI as it exists now will "win out". I think these two technologies will ultimately become so integrated at the vendor level that all products in this space will feature combined SBC and VDI like capabilities.

Your server cost is a bit high, I believe.  Not that it should really matter as this should remain constant for your evaluation.
I believe that it is going to be sooner than 2 years for VDI to overtake Terminal Services in terms of replacing the traditional desktop with a true thin client capable of handling graphics intensive and multimedia applications hosted on either a XenDesktop or VMWare VDI backend server.  This could be the combination of either enhancements to ICA or using protocols such as SPICE as well as development of other protocols.  NEC/WYse and HP RGS are some of the mini workarounds today that address some of this today, but not all.

After reading the original CRN article, Brian's analysis and comments, and additional postings, I have yet to figure out what Virtual / ized Desktop Infrastructure is?

1. Is it a full OS streaming from blades in the Data center? Sounds like the Ardence technology plus AppStream or CPS.

2. Is it VM's delivered onto an all ready running desktop OS? (seems somewhat redundant to existing products and may only benefit a small group of workers).

3. Is it VM's delivered via PxE boot to any end device? (Kind of like a Ghost Multicast from the data center).

I am asking simply because I have never seen a VM of a recent OS with productivity applications installed that is less than 3 GB. I have certainly seen images (Ghost, IMG, ISO, VMDK, VHD) that are larger than 12 GB. Personally, I would not want to be deploying this lovely experience over my network unless every desktop had a 10Gbps NIC.

Please enlighten me. Thanks.



Although this type of debate is useful in discussing the pros and cons of each technology I really don't see why this has to be a shoot out between VDI and SBC. Any CTO with half an ounce of common sense will see the multitude of benefits that each technology brings to the table and plan their infrastructure accordingly.

My own personal opinion is that we won't see a convergence of technologies which provide a single application delivery platform for many many years yet. Until that time comes both VDI and SBC together with server virtualisation and streaming technologies will be the bread and butter of our existence.

Ultimately, desktop virtualisation will die a slow painful death as it is a fat, wasteful technology based on fat bloated OS's. How many CPU cycles, megabytes of memory and disk are wasted in providing duplicate copies of OS's and apps and how much energy does this waste in operation and production?  What percentage of the functionality of Windows XP and Vista will an average user ever need?  Virtualisation is still there in my vision of the future, but the apps sit directly on the virtualisation layer with just a 'skinny  OS' between them and the bare metal, and most of the virtualisation will be handled directly by the hardware. A timeframe for this....3 to 5 years.

So, forget the argument about VDI versus SBC, they are both dinosaurs with a predictable extinction event on the horizon....look to the future, its closer than you think :-)

My $0.02.



The ultimate answer lies in the hardware. If the hardware is powerful enough, who cares what resources the software that runs on it takes? And hardware will get to that stage and software will ALWAYS be trailing hardware. That's just a fact. Now, this matter should not even be an argument to begin with. SBC and VDI serve two completely different purposes and only overlap in a small area of functionality. Why would you need to give a user a full desktop when ALL they user NEEDS is a simple custom made VB app, or a word processor, or a browser window? By the same token, why would you give a user a whole bunch of seamless apps when you can just prepare a desktop with all these apps and just have the user connect to it?

Look at Help4Ctx's post above. Need I say more? He's got it down packed. I have, too, been saying that for years.  It is the hardware what ultimately holds the key to the future. When this happens, all these virtualization technologies will become obsolete, as Help4Ctx above pointed out. 


"The ultimate answer lies in the hardware. If the hardware is powerful enough, who cares what resources the software that runs on it takes? And hardware will get to that stage and software will ALWAYS be trailing hardware"

Because the Earth is warming. Data Centers are at the end of the road in terms of providing power and cooling to increasingly powerfull systems. Most organisations, if they haven't allready, are reducing their server and hardware count, not increasing it. 

I would expect the future lies in cut down resource economical versions of software being able to run in a virtualised hypervisor or a a standard cut down O.S.

Terminal Services as we now know it is dead. Application delivery will morph into the virtualisation layer eventually.


"I believe that it is going to be sooner than 2 years for VDI to overtake Terminal Services"

I agree - in fact probably not long after Citrix get Ardence to work with XenServer, currently it only works with VMWare. This should happen in the next few months though. We've done some testing with Ardence and the Tech Preview of XenDesktop (which is Desktop Server 2.0 basically) and the solution is (or will be) incredibly powerful.

My guess is that, by the end of next year, the start of what could be an(other) enormous growth curve will begin for Citrix. The XenSource acquisition was a massive coup and strategically the best purchase they've ever made. Being in the business of selling this stuff, I personally am licking my lips in anticipation... :-)

Jason, VDI is basically the same as Terminal Server, only instead of using RDP to connect to a Terminal server, you use RDP to connect to a VM which happens to be running on a server with many other VM's.  You shouldn't have to worry about "deploying this lovely experience over my network unless every desktop had a 10Gbps NIC," because the full VM image does not need to be pulled down in order for it to be used.  You can run VDI on slow connections (ISDN 128K), medium connections (cable, DSL), or you can run it at high speeds (T1, T3, etc.), the only difference would be the possible latency time between the client to the server via the RDP connection, which is all actuallity is just doing a screen refresh on the client with all the processing done on the server on the back end.  Think of the dumb terminal days.

Are you suggesting that every desktop computer should ship with just a hypervisor? This would create a similar hardware environment and allow for the virtualized desktop to be deployed from the data center? (OS'es would be licensed in a different manner).

I am still not sure this is a good user experience. Sure, I save Euros at the desktop (power consumption, fewer administrators, etc.) but who owns the process? Desktop guys, network guys, or Data center guys?




Thank you for that clarification. I understand that RDP will be the protocol of choice for many vendors' products. Being the dinosaur that I am and remembering true terminal computing evolving into PC emulation scenarios, I recognize the value of centralized computing models (Even software streaming, which I never knew about until the Citrix / Ardence announcement, is more cost-effective).

With my company investing heavily in infrastructure costs (security, bandwidth, hardware, etc), I would want to see these VM supplied as appliances (pre-configured OS'es in either VHD or VMDK) to make my adoption less costly.

Thanks again,



EXACTLY! Which begs the question, can you really call VDI an application delivery method when it would have to rely on another application devliery method in order for it to become a manageable solution in a large environment?

 There are some really good reasons for moving the desktop environment into the datacenter, but by itself it's not a managable application delivery method.

CTO with a half an ounce of common sense.  haha, what dream world are you living it?  CTOs of large organizations typically are clueless when it comes to common sense.  Their main job is to manage the managers.

It will not remain a constant as VDI hosts need way more memory than a typical terminal server. (OS load for each user, best case 384 per vm).  I have an cost comparison (a number of them) and it turns out to be about 400 Bucks per session on TS with about 1000-1200 per VM session in VDI. I dont use citrix in the calculations since its a TS VS VDI comparison.

there are briforum session videos from 2005 with these numebrs at the end of the presentation for comparison
Ron O 


lol, you may be right, however I've met a few who are so sharp they slice diamonds on a part time basis !!



"Now, this matter should not even be an argument to begin with. SBC and VDI serve two completely different purposes and only overlap in a small area of functionality."

As Brian says: "Of course "VDI" means a lot of different things, depending on who you ask. For the purposes of this article, I'm using the term "VDI" to mean Windows XP or Vista instances running in a data center, with users connecting to them remotely via a protocol like RDP.".

So in what way is VDI "completely different" to SBC, as both involve the user connecting to an O/S running on a server in a data centre, neither require a fat O/S at the client end, and both utilise a thin protocol to connect in either RDP or ICA?


Obviously not, so why would a single solution suit all your users. We too often evaluate technology, and look for simplified answer, rather than asking what users need. Users need applications, not OS's or hardware (these are just a means to an end), and different users have different application needs, which require differing delivery solutions. I strongly believe any decent sized organisation will end up with a mixture of VDI & traditional SBC, along with 'fat clients'. The proportions will depend on the applications, and how users need to access them.

Surely the ultimate answer relies on the software?  If the software is badly written and the ISV won't change it for a customer that wants to 'go thin', virtualised XP desktops are the only alternative.

One advantage VDI has over TS is the need to re-package applications to TS aware(that is provided they were already been packaged to XP) and enable a quicker time to market.

You make a good point but we find combining OS & app streaming INTO the VM gives you roled based access and managed solution you need. I wouldn't bother with AppStream since MS gives away SoftGrid so either Ardence for the OS stream and either embed apps in your image OR use Softgrid to add your apps. We have also seen Wyse WSM handle both the OS streaming AND the app stream piece so in that case you only have 2 pieces; VMware for the hosted desktop and Wyse for both OS and app provisioning. Some good options exist! Good luck!

You might think why would someone bring this topic here. The reason is when you are talking about VDI or TS one does forget that they ,might in future be using office communicator 2007 or later versions of it. Only presence and IM work in such environment. THe problem comes when you want to strem video and audio. Thats unsupported and can't be achieved. SO while considering sloutions like this please ensure that you are not using OC 2007. An additional information as we are talking about virtualization. OCS 2007 roles on Virtual server /VMWARE is not supported. It may or may not work. Please refer to the document: