If you're doing VDI for less than 500 users, you should use DaaS

People often ask me for my opinion on which desktop virtualization product is "best." Of course that answer depends on many factors for a given scenario, but one of the most important questions to before deciding is, "How many users are you building this VDI for?

People often ask me for my opinion on which desktop virtualization product is "best." Of course that answer depends on many factors for a given scenario, but one of the most important questions to before deciding is, "How many users are you building this VDI for?" I'm surprised by how often I hear answers like "50" or "100." And when I hear that people are building smaller VDI environments, I can't help but think that they can't do it better, cheaper, or more securely than a cloud desktop / desktop-as-a-service (DaaS) at that scale. So just go to one of the cloud providers and be done!

The reality is that even in 2013, VDI is complex. Sure, there are hardware makers like Nutanix and Simplivity who combine compute and virtual storage into single appliances, and there are virtual storage providers like GreenBytes and Atlantis who let you run advanced virtual storage on the same hardware you use to host your desktops, but that doesn't negate the fact that VDI is some complicated stuff!

If you think about it in terms of effort-to-costs, I've often talked about how you can build some decent VDI now for $500 per user. But that's based around a fair amount of scale. So if you're building a small environment, I'd consider bagging it and just going and buy it from someone who knows what they're doing and can get the huge economies of scale in their environment with savings they can pass on to you.

I mean even if you could build your own VDI for $500 a user, if you can buy VDI for $30 per user per month, you're talking about a 17 month break even point (which is halfway to your hardware refresh anyway), and all that time you wouldn't have to deal with swapping out hard drives, upgrading users, or managing servers.

if you're not following this logic, think about it in the context of other areas of IT. I mean look at Gmail. What is it that makes a private company think they can run an email service more cheaply, more reliably, and more securely than Google? People say, "Yeah, but that means our mail is in the cloud. We need to know where our data is?" Really? Why. For security reasons? What do you think is more secure—Google's servers and datacenters or your office? People say, "But Google doesn't have an SLA!" Um, yeah, that's because they're Google. They don't need an SLA. When Gmail is down it's in the news, and they have about 100x more people than you have all scrambling to fix the problem. (And many of them are the people who wrote the email product!) When your email goes down, you have Ray, your Exchange-certified sysadmin who you've paged several times.)

So when it comes to VDI, I don't know what the exact tipping point is. 50 users? 500? 1,000? Let me tell you, it's big. The company I work for (TechTarget) has about 700 employees. If we decided that VDI was the right solution for the enterprise desktop for our users, there's no way we could do it better than a Desktone or tuCloud. If we went out to a cloud provider, we could spend all of our energy focusing on designing our desktops. We wouldn't have to think about reference architectures or IOPS or LoginVSI metrics or design best practices. We;d just say, "we'll take 700 please" and we'd be on our way. Done.

So regardless of where the exact point it, every company has a number of desktops under which it just doesn't make sense for them to do VDI on their own. And I'll bet if we really dig into it, that number's a lot higher than many of us think.

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

Brian, I think you are leaving out a critical consideration.  As you often say, it's not about the desktop OS it's about the applications and the user experience.  This is something I fully agree with and in the smaller use cases you describe you must consider the location of data and applications.  If you move the desktop into the cloud, where does that file server go?  That one may be simple, but where does that legacy client/server application infrastructure go and how are they accessed by the cloud desktops?  

You could easily end up with a worse user experience by separating the desktop from the application/data over a WAN/Internet link.  The easy answer is to move all your applications to a SaaS model but that's just not reality for everything.  There are services out there which allow you to move not only your desktops but also any other legacy physical or virtual server/application infrastructure into a common cloud so that desktop to application/data link remains at robust LAN speeds.  Something to take into serious consideration if you are contemplating making this sort of move.

I, like you, believe that XenApp or RDSH is more than suitable for these smaller user counts in most cases and can be done very cost effectively so don't forget about that.  Keep it Simple!! :)


Another angle I'm seeing is that the applications we were once putting our virtual desktops closer to in our datacenters, are quickly moving into the cloud as well.  

Not just web-based PaaS apps, but all those nasty ADO/SQL backed VB6 grid-control Win32 client-server apps that love them some Windows run-times.


Chris raises a great point - why not go with one of the thousands of service providers that offers XA as a Service today? Remember that economies of scale are harder for the SP to provide for VDI due to limitations on what can be shared across customers. They might get buying power, but unless you can fully load one or more servers with VDI your better option is an RDSH based solution like you can get from XenApp or XenDesktop.


I agree with the premise... Citrix Service Providers have reported average DaaS deals in the 100-200 seat range, though many above 500 seats. Many economies are achieved via shared infrastructure, per Calvin's note above.  

Interestingly, deals > 500 seats are sometimes showing up a bit more expensive since the enterprises generally want some level of customization & special maintenance...


Plug-n-play (your credit card) for rapid time to value is making more and more sense for business needs. Education is a perfect example where hosting the desktop, giving students access, and the institution focuses on their core competencies of learning, which makes good sense. The app inventory is stable and the turn-over is regular. As-a-service providers now offering hybrid local/cloud architectures now deal with final mile and even data location issues. Yes, the space is still evolving (improving) as is the cloud architecture in general… which is headed to hybrid as well. On a separate note, the $500 per user cost needs some clarification. The physical parts of VDI are now down under $200/user. Consolidated infrastructure and storage optimization has now put the acquisition cost much lower (and yes, it can be a persistent user too).


@Brian I think you are incorrect to assume than small user base means not complex. If it's simple use cases, then sure DaaS providers or XenApp/RDS may be the best choice. I've worked with a few people in small shops where the user demands are very diverse and VDI there makes sense, but is very diverse on the app side. This is where the DaaS providers just don't cut it. A service provider who knows your workplace is more likely to succeed, else Build Your Own.


@Brian Agreed :)

But Appdetective makes a very valid point, DaaS providers are just as averse to dealing with lots of funky apps as the rest of the EUC space.


@AppDetective Amen, Small user base != low complexity.

DaaS providers are fine for Microsoft office and one or two challenging applications, but as the environment starts to require more features and work arounds the DaaS solution becomes less flexible and forgiving than an on premise installation.

I've seen this lack of flexibility on more than one occasion leave the customer faced with a 'computer says no' response from the DaaS provider canning the project.


@Calvin  I know you Citrites would love it if everyone just used XenApp and Terminal Services and whilst I agree with you that the restrictions placed on DaaS unfairly favour RDS/Xa to the point it often makes more sense to do Xa/RDS, I think you miss the point.

Nobody wants to use a decades old legacy tech to host their desktops in the cloud, Xa/TS is completely unsuitable for hosting large amounts of desktop clouds as most of your HVD provider partners are currently discovering, the limp mule doesn't scale well over tens of thousands of desktops.

Then there is the way most HVD providers architect their solutions is for density, which is a ticking time bomb security issue,  All those users sharing the same core, the same dom zero, all packed like sardines onto a persistent server image, its not going to work see, no proper separation of personalities across a multi-tenant infra.  Insanity it is.

RDS/Xa is very obviously going the way of the dinosaurs and if the best reason you have to deploy it is because 'VDI isn't allowed to do stuff', then you will never get it.

These days its much easier to host small amounts of desktops using the free version of tuCloud DaaS Engine on your own metal, we made it ridiculously easy to deploy desktops and we give it away for free in order to disrupt the SMB space whilst we focus on the enterprise.

Its REALLY easy and cheap to DIY DaaS if you have a few desktops, just google tuCloud DaaS Engine and download our baby, 20 minutes later you have DaaS.


I agree with @Guise: XA/TS is not the answer. Most desktop environments are too complex to be replaced with an XA based solution. The XA/TS environment works fine if a service provider is trying to deliver Microsoft Office as a Service. But not replace the full capabilities of almost any desktop.

I agree with @Brian: Anybody with less than 500 desktops should definitely think DaaS first. There is a tremendous amount of complexity in deploying and managing a new set of infrastructure - servers, storage, networks. These are not easily available skill-sets and when things go wrong its not one desktop that fails, but 500!

I think it might be time even for larger deployments to think DaaS first. The cost/complexity arguments may be even truer for them.