Is virtual desktop infrastructure technology a passing trend or a long-term solution to our desktop needs? We'll explore 430 survey responses.
Virtual desktop infrastructure technology is gaining popularity fast: Of 430 respondents to our July 2010 InformationWeek Analytics Desktop Virtualization Survey, 77% are either actively using or testing VDI (42%) or assessing its benefits (35%). Most implementers do limit VDI usage based on category, with adoption ranging from 68% of corporate office workers down to a mere 3% of media production workers. That makes sense given VDI's current limitations in the areas of heavy graphical load and remote connectivity.
Intel is considering bucking that trend, however, evaluating VDI for a range of applications--including those requiring dedicated GPUs for higher-end graphics. "Intel has an 81% notebook to 19% desktop mix, so we are very mobile and place a high value on the productivity we gain from that mobility," says John Dunlop, enterprise architect with Intel's internal IT group. "This includes the ability to work offline and integrate graphics and multimedia-intensive collaboration tools that have historically been challenging to scale with VDI."
Dunlop recently completed a formal VDI evaluation and total cost of ownership analysis, including testing capabilities to support multimedia collaboration, unified communications, and training without an undue hit on shared LAN segments.
For our poll respondents, the impetus to explore VDI ranges from a way to finally get off the fat-client treadmill to a stopgap until workers' desktops morph into something else. A few see VDI as a plot by OS and application vendors to charge us double for licenses.Whatever their take, respondents react strongly to the technology.
The Cost Factor
"The up-front licensing costs, along with the perception that users still need PCs, and PCs dropping in price, pretty well nixed the concept," says the technical architecture manager for a large retailer. Adds another: "Frankly, with SaaS and other Web apps coming online, I look at desktop virtualization as a Band-Aid to the desktop management problem." And from a public sector IT director: "Microsoft is the biggest problem and cost with VDI. Why should I have to pay another fee?"
It's a good question, and one respondents are wrestling with. Sixty-two percent of those using, piloting, testing, or assessing VDI predict that application and/or OS software licensing will be either the same or more expensive than current fat-desktop pricing and terms. Few expect to save any money, a reality borne out by Dunlop's testing.
"Current implementations of VDI may increase TCO for primary device replacement scenarios," he says. "Large enterprises may not be able to recapture VDI investments--infrastructure, licenses--from improvements in centralized management if clients are already well managed. Instead, we found most of the value in VDI to be in augmenting the user's primary client device." Dunlop cites software licensing and SAN costs as the most significant drags on TCO.
So if it's not an expectation of savings that's driving interest in VDI, what is?
Security and easier maintenance are the top expected advantages among our survey respondents.
In concept, VDI is based on server virtualization, with many of the same benefits and drawbacks. Just like server virtualization, VDI serves desktops to users by running multiple virtual instances on a standardized hardware infrastructure in the data center. Desktop workloads are stored on a common storage network. Groups of desktop computers, or pools, are generated from a single master image and deployed automatically by the virtual infrastructure software.
Because the machines are deployed from a single image and are--initially, at least--identical, the common user base receives a highly standardized and structured desktop experience. Management is improved because changes made at the base image can be propagated right down the line to all desktops in the pool, while still maintaining user customizations. Meanwhile, security is enhanced because data and applications can be tucked away in the data center. End users simply connect to the host server farm, which may be physically located in the same building or a country away.
Shades Of Problems Past
Unfortunately, VDI also injects an additional layer of complexity into the machine provisioning and maintenance process, especially if, like Dunlop, you plan to support a wide range of devices and demanding applications. Just as with traditional utilities, the master image must be flawlessly prepared and thoroughly tested, or problems will be multiplied. The dynamic nature of desktop maintenance also presents problems for some legacy applications, so IT must use care when creating user environments. Since virtual desktops will be running in the data center with direct server access, it's usually advantageous to warehouse user data, like desktop contents and document folders, on storage servers instead of as differential change files for individual desktop images.
As we discuss in our full research report on virtual desktop infrastructure, available to InformationWeek Analytics subscribers at informationweek.com/ analytics/vdi2010, a set of best practices is emerging for companies that want to maximize savings and security while minimizing user discontent. For example, Intel's Dunlop stresses not conflating VDI and thin clients; in many cases, the true value of VDI may be in augmenting the existing end user setup.
Still, respondents are clearly bullish: Within 18 months, more than a quarter of respondents using, piloting, testing, or assessing VDI expect to have between 26% and 50% of their desktops virtualized; an additional 16% say they'll have up to 75% virtualized. One area they're not so excited about is outsourced VDI. Just 14% say they'd definitely consider such services, while 39% gave us a definitive no for an answer. To decide what role VDI should play in your company:
>> Set realistic goals. VDI shines at companies with large, homogeneous user pools; in highly regulated environments; and in private cloud scenarios where desktops are served to remote users, provided that graphics aren't a major factor. Intel's Dunlop stresses that IT must understand usage requirements right down to the application characterization level and recommends evaluating graphics requirements from a performance and latency perspective.
>> Gather a team. VDI success requires a nontraditional technology skill set, and maintenance savings will be realized only by replacing entry-level desktop support with fewer, more highly skilled virtualization pros.
>> Limit pools. Begin with a well-documented inventory of prospective VDI users that identifies the types of applications various roles require and how they're used. Break employees into as few groups as possible. Customization raises costs.
>> Put real numbers to licensing requirements. Licensing for virtualization platforms can get very expensive, very fast. Although you should only have to make the investment once with software assurance, it can be hard to sell to management.
>> Develop criteria to measure return on investment. Key ROI metrics include maintenance savings, faster and more effective desktop provisioning, ability to consolidate infrastructures that couldn't previously be centralized, ability to support more devices, and compliance and security management improvements. What's critical is identifying which areas are most important to you, and use these to balance expenditures.
>> Select partners wisely. Vendors will continue to improve VDI feature sets to rectify real and perceived technology deficits. However, how well companies and their application vendors work together will determine whether virtual desktops mature to fulfill IT's ongoing quest for smooth and secure desktop operations.
Server Market SplitsvilleJust because the server market's in the doldrums doesn't mean innovation has ceased. Far from it -- server technology is enjoying the biggest renaissance since the dawn of x86 systems. But the primary driver is now service providers, not enterprises.