The savings that flow from server virtualization are well known and accepted. The potential savings on the client side, I believe, are even greater. But that premise is much less widely accepted, and even less frequently implemented. Perhaps virtualizing end users one application at a time is the way to go.When it comes to virtualizing clients, IT managers take a step back and hunker down. Even if they can find the right connection broker and identity and authentication servers, there's still the issue of providing a satisfactory end-user experience. And those pesky end users sure let you know about it when the experience isn't satisfactory. If there's no clear solution, who needs to volunteer for that headache?

So client-side virtualization thus far has been a nonstarter. There are still just too many issues to resolve. If you try to store everyone's individual desktop as a virtualized file on a central server, then you will have increased storage costs and maybe, or then again maybe not, decreased client costs. Who wishes to be first in demonstrating how planned savings evaporate? Several vendors claim a complete solution to this problem but lack a list of high-powered customers who've adopted it.

Then there's the issue of not all end users are alike, so you'll need a variety of virtualized desktops, and make sure you always know how each one evolved, so you can re-create it. There are partial solutions -- some would claim whole solutions -- to these problems, but the world isn't flocking to them either. But there may be a partial solution that yields some savings while giving both implementers and end users needed experience in client virtualization.

End user applications can be virtualized one by one on central servers and made available, without virtualizing the whole desktop. A lesser-used application, such as Microsoft Project, might offer a test case of what can be accomplished and where the cost savings can actually be captured.

Or a set of applications could be aimed at a particular end-user group whose needs were more basic and more assured of being met than others. A test of the capacity of the corporate network to deliver virtualized end-user services in a manner that mimics the performance of the desktop would yield valuable information.

In a declining economy, these small initiatives may pale against the CFO's next demand for real IT savings. But the sooner IT takes some stab at desktop virtualization, the better armed it will be at confronting the task when there's no longer a choice on whether it's going to happen.