Blog post

Seamless Application from a Virtual Desktop: If only this made financial sense

By Gunnar Berger | October 11, 2013 | 7 Comments

I take a lot of heat about my view of virtual desktops. I don’t think a week goes by that I’m not involved in some public debate about which should I use RDS or VDI. In fact a large part of my job is in helping our clients make that decision. The long and the short of it is (and this is a fact most people will agree with) you use either of these depending on your needs and use case. The devil is in the details which is why I do a lot of research about this. There is one detail that I don’t talk about too much and that’s my view on RDS over the long haul verse VDI over the long haul.

Today the simple fact is RDS is cheaper than VDI. There are numerous reasons for this; one is the license difference between the RDS CAL and the VDA license required for VDI. RDS also has a SPLA license which makes it easier and cost effective to host RDS applications and desktop from the cloud. Another reason is that RDS scales better than VDI, however I argue that at some point this will be a non-issue for the majority of organizations. There is a saturation point of how many desktops you want on a piece of hardware before you run to high of a risk (too many eggs in one basket), I am a fan of 75-125 users per server myself (the lower the better). So I don’t dig into the scale argument too much, yes RDS will always scale better, I typically only care about this in large environments. The second argument on the cost of licensing is one I’ve been vocal about (#FixVDA) and I’ll spare you another blog on that.

So if you will,  imagine a world where RDS and VDI are very cost competitive, in this world I would lean very heavily on VDI and this drawing is my reason for that:


In this drawing I present a list of applications installed on a client OS. The major difference I’m showing is that the application is being seamlessly delivered from that same Client OS. This is a feature that is already available from some vendors but not commonly used. I see some inherent advantages to this:

  • Flexibility in application deployment: From an application stand point this model is exactly how a typical user uses a desktop: some of applications may be installed by corporate IT, others may be user installed. The great news is in this scenario there won’t be any application compatibility issues because most Windows apps are designed to run on Windows client OS. Also meetings that require a special WebEx, GoToMeeting, software agent will work just fine because we are enabling users to install their own apps, something you’d never do on RDS.
  • Application Compatibility: Since its running on a client OS it solves any application compatibility complaints.
  • One Platform: I’m enabling the sending of a seamless application from a client OS which means IT organizations don’t have to have separate infrastructure and management for delivering RDS or VDI, we just use VDI for everything.
  • Follow-me-Application: Since this platform is completely based on persistent desktops a user can be working on something on their desktop/thin client and walk way, flip to a mobile device and pick up where they left off (ideally with a seamless app).
  • Better UX over Mobile: Another VERY important component of this platform is skinning the application so it works better on a mobile device, most VDI vendors are already working on this and I just wrote a report about it. This platform would make it so if a user was using an application and saved it to their desktop it would always be there because its just their desktop (of course by now I’d rather you use a mobile file sync solution but that’s a story for another blog).

The point with this madness is that to a user this will be their desktop, they customize it how they want. Admins will be able to take advantage of management technologies (such as application layering) to simplify the management of these devices. Using a fully persistent model makes it even easier to deploy and understand (low learning curve) for organizations that don’t have the expertise needed for traditional VDI.

This may seem like a crazy concept, and it is, right now it doesn’t make any financial sense to do this. If you want to play with the idea I believe XenDesktop 7 can do this however, I haven’t tested it out myself yet. For now, you and I still need to talk about when you should use RDS and when you should use VDI (or when you should use both) but I like this potential future and I really do hope we get to a place where RDS vs VDI is just a decision based on user experience not based on cost.

Gunnar Berger

Follow me on twitter: @gunnarwb

Disclaimer: As people poke holes in this solution I will change this blog to create a stronger position. I put it out in the public because I like the discussion this brings on twitter and in the comments. As is true in all blogs this subject is just something on my mind it is not an official Gartner stance.

Leave a Comment


  • Issa Banna says:

    Text me 4693584618

  • You have summarized the solution that the XenDesktop team has built over the last five years. Spot on.

    The cost argument is important for an employee who spends the majority of their time sitting at a desk at work.

    In many cases, the cost argument is not important. For example, if you want to deliver an enterprise desktop to an offshore employee, contractor, consultant, remote employee, or a mobile employee there is really no better alternative than a virtual desktop.

    Isn’t the remote/mobile/contractor/offshore/consultant use case almost 50% of a typical organization? So desktop virtualization is definitely not a 100% solution, yet, but I believe that its a 30-50% solution.

  • Stu Downes says:

    Hi Gunnar, I suspect your model is good for some use cases. There are important elements of user experience we need to consider in this concept. The main use case I can see on the apps side is where:

    (i) the data the app accesses is so sensitive you need to keep it in the datacenter, and
    (ii) the app itself does not have a native client for the endpoint in question.

    If we take most productivity apps my view is that on the native device there is either an app from the enterprise vendor or an app from a 3rd party which would offer a better user experience; combined with MAM/MDM/file sync a better user experience results.

    For me its about the context of what application is required on the device and the sensitivity of the data/compatibility of that application with the local client OS.

    I would suspect this would be a step 1 on a 2 step strategy to ensure an application is available on a Windows or non-Windows endpoint.

    As you say though the complexity of VDA/Companion Licences/Win+SA/On Network/Off Network needs reducing beforehand.

  • G. Simonson says:

    I’d like to point out some general observations regarding VDI vs RDS cost. In a large financial institution I am working for there are 30K XenApp users supported by 30 second and third level support engineers, and 25K VDI desktops supported by a team of 5. In both cases thousands of applications are in use, using the same delivery methods. If we consider this XenApp support complexity (requiring much larger number of FTEs to support) and recent developments in VDI storage architectures (i.e. flash and deduplication), this could change the cost balance significantly, offsetting bigger VDI hardware investments. Actually, Gunnar, you are in a good position to research this, making some conversations with your clients and verifying if this trend could be true.

  • Gunnar Berger says:

    G. Simonson,

    Your observations are in contrast to our observations. RDS is typically easier to manage than VDI. Although they overlap in many ways, it’s the storage aspect of VDI that increases the complexity. This is getting easier but storage remains an increase in both cost and management.


  • G. Simonson says:

    Actually I failed to realise and mention that application support on VDI is done by an additional support team, and often in a desktop by desktop approach, while with SBC case it is centralised for all users. So this makes SBC (still) look better from the support cost perspective. I wonder if application virtualisation together with profile management could alleviate this?

  • CyberlinkAsp says:

    Your services are highly useful for our business.