Over the last two years, Intel commissioned a study on how companies were delivering virtualization to end users. It's one of the few indicators of where this confused segment is headed. Several approaches are still on the table, but the fastest growing one is where the virtual machine resides on the end user's PC."Client-based virtual containers show the strongest, expected growth," the second study concluded after 396 online interviews were completed in November 2008.
How it reached that conclusion is not clear to me, because it didn't offer that category as a choice in the survey the year before. In the 2007 survey, the concept of virtualized desktop was based on server-originated virtual desktops only .
Nevertheless, the study was done in partnership with Citrix Systems, which, along with VMware, began offering client-based virtual containers in its product line two years ago. Citrix XenDesktop is now in its third version; VMware's ACE , release 2.5.
Both are the type of desktop where a user's preferred operating system and applications are combined in a container, or cordoned-off set of virtualized files, with access to the PC's CPU and other resources in a defined manner. For example, a user might only be able to access certain areas of the hard drive. Under this approach, if the same desktop served different workers on alternating days, then the each could have a distinct desktop virtual machine without invading the data or disrupting the operations of the other.
Such a "client-based virtual container" will be maintained by a central server, which periodically checks the desktop to patch the operating system or update applications. But unlike other forms of desktop virtualization, it's not dependent on being connected by the network to that server. Disconnect the laptop, for example, from the network and the virtual desktop still functions enroute to its next networked destination.
That use case may provide the key to how desktop virtualization will evolve. The Intel/Citrix survey also showed strong interest in the "OS streaming" model, where both the applications and OS are sent down by a central server at the start of each session. This approach has strong IT management and security features, but doesn't work for people who are road warriors or mobile workers. If they're disconnected, there's no central server to get them started.
The other types of desktop virtualization described in the survey were: "terminal services," where a dumb terminal or in this survey's parlance, a "networked terminal," is tied to Microsoft Terminal Services or Citrix Presentation Server; "virtual hosted desktop," or a client operating system and applications hosted on a central server, with the presentation sent to a thin client or PC; "application streaming," or end user applications stored on a central server but streamed down to the client for execution there; and the previously mentioned "client-based virtual container" and "OS streaming."
Of these types of desktop virtualization, all are deployed primarily on standard desktop PCs or laptops. Thin clients only marginally fit into the picture, although that may reflect the existing infrastructure and not future investment. Over time, where end users are fixed in place, the thin client appears to be a money-saving virtual desktop machine.
One other form of virtual desktop was included, and that's the PC blade, where a blade server in the data center serves as a remote PC to a networked end user or, more likely, users. Twenty-six percent of IT managers reported using blade PCs in 2007; 29% in 2008. But they most likely represent a small fraction, the power segment, of the total end user population.
Given that an enterprise may employ more than one of the above models, there's no simple set of figures adding up to 100% to indicate how much of the total desktop market each one represents. OS streaming, for example, was in use by 15% of 705 IT managers interviewed in 2007; 20% of 396 IT managers in 2008. Lowly terminal services were reported in decline: 64% in 2007; 54% in 2008.
Application streaming was in use in 30% of cases in 2007; 31% in 2008. In other words, Citrix acquired XenSource as it sensed its Presentation Server application-streaming product was leveling off a growth curve. Other forms of desktop virtualization were going to take over, and it maneuvered accordingly.
But like I said, no reference was made in 2007 to the strongest growth model, client-based virtual containers. The study showed them used by 20% of IT managers in 2008 and projects that use will jump to 29% in 2009, an increase of 45%.
I think this study tries to show, without a lot of evidence, that end users want their virtual desktops to duplicate all the advantages of their physical desktops. The large percentage of PC users who are periodically detached from their cubicles or offices and working somewhere else practically dictates that the client-based virtual container will be the model that serves many end users in the future.
That may be just a desired conclusion by Intel and Citrix; they're both heavily invested in that model. But it has a resonance to it that may point the way to the future. End user preferences are going to be much more in play on this front than in data center server virtualization.