Desktop virtualization

As a follow up to yesterday’s blog on the hypervisor revolution I wanted to spend a few cycles talking about the desktop. For the last 30 years we have become accustomed to personally managing our desktop OS. From business to consumer and back to business desktops our sense of ownership and entitlement has come at a very high cost. The cost of management, security, patching, corruptions / conflicts and ultimately the cost of the hardware upgrade (or replacement) necessary to support the next / latest version of the software we need to get our job done.

Now the idea of running your desktop OS or at least a subset of applications from the server is nothing new. The first desktops were dumb terminals running mainframe sessions. The evolution to thin clients was somewhat revolutionary in that you had your own GUI-based desktop image. However because the network wasn’t fast or efficient enough the thin client solutions were forced to make compromises that in most cases negated the benefits of moving off a traditional PC in the first place. You still had an OS image on the desktop to manage, patch and secure. In addition the customer often had a set of legacy peripherals that they were forced to scrap as they weren’t supported by the thin client. In addition many solutions had architectural issues -- single points of failure -- or unique management interfaces that they had to learn and certify. Often it was more challenging than simply sticking with the tried and true. Now, I’m generalizing and for simplicity not even dealing with application virtualization as an option. In addition many customers moved to a bare bones Windows virtualization solution from Microsoft called Terminal Services. This functions fine unless the user needs some basic things like support for sound. ;)

Which brings us back to the hypervisor. The wonderful thing about virtualization is that you can run any OS (x86-based that is) on the hypervisor on a server and deliver it via the network. The benefits of this model -- outside of centralized imaging, managing, patching, service and support -- is that you can optimize the network to a point that minimizes the technological footprint of the client yet delivers almost full desktop capabilities to an end user. Thus, in the last two years the introduction of the Zero Client. (I take some credit for that concept!) Yes a Zero Client may have some small firmware or bootware but for all intents and purposes we’re talking about a client side box that provides ports for network and peripheral access. (A place to plug in your monitor, mouse and keyboard.) There are a number of Zero Clients on the market today led in large part by Teradici’s OEM suppliers (such as ClearCube), Pano Logic and nComputing. Now for the initiated I don’t want to get in a discussion on whether Teradici or nComputing are indeed desktop virtualization plays at all considering their unique architectures or whether they (today) utilize a hypervisor backend solution -- in those two cases they don’t.

The background was important for my point: while this is revolutionary it ultimately doesn’t matter!

None of this matters for two reasons: commoditization and cheap alternatives.

1. The three most significant components of desktop virtualization and its alternatives are: connection brokers (software that authenticates and connects remote users), hypervisors (disaggregators of the OS) and protocols (optimized network connections). Connection brokers were commoditized two years ago. You want a connection broker you call any vendor in the space and they’ll give it to you for free. There are no companies specializing in connection brokers left. VMware, Microsoft, Citrix and all the smaller vendors provide the connection broker at no cost. The hypervisor is being commoditized as I write this blog. VMware tried to sell ESX for $999 and Microsoft countered with their hypervisor, Hyper-V, for $28. Now you can get either for free. Xen the open source hypervisor has always been available for free. Sun Microsystem’s Xen derivative, XVM is also available for free. The highest potential hypervisor, KVM will begin shipping as part of Red Hat any quarter now as well. Their business, as VMware’s and Microsoft’s is to build a business on management tools and not on the underlying technology. Lastly there is the protocol. This is where most vendors in the space are focused on creating added value and differentiation. However with Microsoft working diligently to significantly upgrade RDP (provided for free) and VMware partnering with Teradici to distribute PCOIP later this year I predict that soon the protocol will also be free. This leaves tertiary players such as Pano Logic, Sun Microsystems, Wyse and even Citrix with their robust ICA stuck between a rock and a hard place to continue to drive customer’s to pay for little additional value add.
2. Devices ultimately are unimportant because the value of any server-centric desktop solution should be to deliver an optimized experience to ANY device. Yes there are benefits to a Zero Client but the installed base of PC’s, netbooks, and most importantly cell phones is vast and each has a very unique value that will never be eliminated by thin clients or zero clients. Zero clients in particular have other challenges: since they have no state they need a secure network connection. This eliminates their ability to connect outside the intranet (at least easily and cheaply). This is yet another major obstacle.

The net of this discussion is that there is no doubt that the desktop computing platform will alter dramatically from the traditional PC most of us use today (and that includes many existing laptop users). Will it mostly look like a Thin or Zero client? Not likely.