At Catalyst North America last week I talked about the concept of a virtual desktop NAS and wanted to provide a little more detail. When most folks think about desktop virtualization, a deployment model based on virtual infrastructure is often what first comes to mind. Both VMware View and Citrix XenDesktop, for example, rely on server virtualization for their back end infrastructure, which requires hypervisors (e.g., ESX, Xen, Hyper-V), physical servers, networked storage, and associated Ethernet and storage connectivity. Of course, that’s not to mention the software needed to manage the virtual infrastructure as well. Don’t get me wrong. I’m not trying to bash the existing desktop virtualization deployment models of vendors such as VMware, Citrix, and Microsoft. In fact, I’ve recommended desktop and application virtualization solutions from all of those vendors on several occasions.
That all being said, I’ve also talked with many clients who are simply priced out of desktop virtualization by the infrastructure requirements and related up front capital expenses. This is especially true with small and medium businesses, which is where I see desktop virtualization appliances as having an opportunity to fill a major void. Consider this example.
Sure – the illustration is overly simple, but that’s my point. Suppose you could drop in a single box that serves up virtual desktops, while providing much of the centralized image management you want to get out of desktop virtualization? That’s how I see a virtual desktop server appliance, or virtual desktop NAS (call it what you like… I’m not a marketing guy) playing out. Virtual infrastructure in a box is nothing new – and this model could certainly go in that direction with ESX, Xen, Hyper-V, or KVM hypervisors – but I see another option that involves leaving the virtual infrastructure behind. Why not serve up virtual desktops over CIFS or NFS? iSCSI would be an option too. That means all you really need is a basic Windows or Linux file server and you’re there. Microsoft made a reasonable dent in the NAS space with Windows Storage Server and there’s no reason to think that a desktop virtualization appliance couldn’t have a greater impact.
To see what I’m getting at, consider this example. Take a Windows Server 2008 host and drop a virtualization product like RingCube’s vDesk on it. vDesk can serve up virtual desktops over CIFS to Windows client endpoints. The desktops execute locally on the endpoint device where a generic Windows OS is present, such as XP. The virtual desktop container includes a virtualized Windows shell (including domain membership), applications, and the user environment. Virtual desktop updates are centrally managed on the Windows server. Since the desktops are served up over CIFS, you could use existing Windows file replication tools to replicate the environment to an alternate site. So a typical VAR might drop in a local desktop appliance to a small office, and configure it to replicate to the VAR’s data center for backup and disaster recovery. Large enterprises could use a similar model. RingCube is one example, and other vendors such as NComputing, MokaFive, and Virtual Bridges are enablers for the virtual desktop NAS as well.
Of course, you need reliability, and you can get it without clustering and shared storage by using one of a myriad of file server replication utilities. Another alternative would be to package the appliance using a Stratus FT server. The server OEMs did well with Windows Storage Server, but we’re looking at a desktop virtualization market that is larger by several orders of magnitude. I don’t think it would be difficult for vendors such as Dell, HP, IBM, Sun, or Fujitsu to package an appliance that included an OS and one of the desktop virtualization packages that I mentioned. Heck, even Microsoft could jump in and create a platform similar to Windows Storage Server – perhaps Windows Desktop Server or something similar. You get the idea. Again – I’m not here to say that the virtual desktop NAS will overtake the Citrix or VMware solutions, but like we’ve seen with NAS and storage, I think there could be a considerable market for the virtual desktop NAS.
What do you think?
#1 by Doug - August 5th, 2009 at 15:34
If I can take my vendor hat off for a second…
This approach is long over due and definitely an exciting innovation because of its simplicity. There are many customers who get “sticker shock” when considering VDI. As VMware’s TCO calculator shows, it’s $3.3M USD for 2500 users using View. That’s simply too expensive for most companies.
NAS vendors and File Server vendors (e.g. Dell, HP, NetApp, Sun, Stratus, etc.) can greatly simplify and drive down the capital cost of desktop virtualization particularly if several hundred, even thousands, of users can be served from a single file server or NAS appliance.
As the desktop virtualization market matures, the upfront capital costs must decrease so innovations like this one will help blaze the trail.
#2 by zAchEr - August 6th, 2009 at 01:11
From a consumer’s point of view, healthy competition will always go down better for the consumers so if vendors start to jump in to this boat and offer a more cash-friendly alternative to the existing infrastructure – probably with a reasonable decrease in capability – then like what Doug said, it might lead the way for the desktop virtualization industry.
#3 by Bert - August 7th, 2009 at 11:53
As the desktops execute locally, the thin client and repurposed PC should better be left out of your diagram I guess. Then there’s also the host O/S to be managed on the endpoint devices (security patches, updates). It’s not easy to do this from a NAS appliance :-)
All in all, this may limit the use cases as compared to VDI.
Btw, the Virtual Bridges link is pointing to the wrong URL and should probably be http://vbridges.com/wp/
#4 by Chris - August 7th, 2009 at 14:30
Good points, Bert. I agree on RingCube, but NComputing, Virtual Bridges, or MokaFive could make use of repurposed PCs. Thinking out a bit, the endpoint could be a Netbook running a small Windows kernel and the NAS could stream the user environment and apps. Such a model would favor MS since it would protect both server and desktop revenue streams.
Your right that this model is not for everyone, but I think there are cases where it would make sense.
#5 by Lon - August 19th, 2009 at 10:20
I have long been a prponent for this type of DV. It just makes sense for many small-med companies who a) don’t have multiple servers in need ov virtualization, b)don’t have the staff to maintain additional platforms (e.g. hypervisors), or c) don’t have the $$$. We fall into all three of these categories. I just want to be able to start virtualizing desktops becuase that is where the biggest return is for us. We have one server (Exchange, File services, Internet) and <100 users. I would like to find an appliance that would allow using the existing desktops to “connect” using a client (could be browser-based) and also allow thin-client connection. Connection via RDP and GoToMyPC and other remote desktop solutions would be icing on the cake! Someone will introduce this “miracle appliance” soon. And when they do, they will find a hungry market for it.
#6 by Ray Love - September 8th, 2009 at 18:53
I do think that Desktop Virtualization is a great concept and can become a practical solution.
I recently went to a sales luncheon with four sponsors. DoubleTake was one that presented a product that seems to fit in the Desktop Virtualization space that you are describing.
I think their product is called Flex. Although I don’t have specifics, it sounds like a similar concept.
I am confident that the desktop virtualization going to continue be a very competitive field from the players jumping in.
Being a Small Business, we will likely have to wait out the solutions to drop to our cost level.
But the creativeness of solutions like this are bound to change how companies deploy their desktops and applications in the future.
As all solutions, they also have different affects on the infrastructure.
Seems like some drive more hardware into the datacenter, while others seem to redistribute the datacenter load back on the desktops.