|Virtualization: Boon In The Datacenter, Bust On The Desktop |
1/24/2012 @ 9:17AM
It may be startling news, but not all virtualization technologies are created equal.
In the past decade, CIOs and other IT managers had a fantastic experience using virtualization to consolidate their servers, sometimes ending up with one-tenth the number of machines. The fully virtualized servers were bigger, but they also ran near capacity, which the old physical servers rarely did. Costs were reduced, efficiencies gained, and service levels increased.
After such a great experience, who could blame IT people for thinking they had found a panacea for every ill? Unfortunately, while virtualizing servers is a no brainer, doing the same with desktops just leads to a sinkhole of complexity, unforeseen costs, and little return on investment.
And CIOs are catching on. Four or five years ago, these empowered employees had just come off the most successful IT implementation in history: using VMware to turn thousands of servers into hundreds. They wanted to do the same thing with unruly desktops.
The primary vendors of virtualized desktop infrastructure (VDI) — VMware, Citrix, and Microsoft — told everyone that VDI would reduce ownership and security costs. It made sense on paper. With everything kept in the datacenter, IT could update a single copy of Microsoft Office and push it out to all the thin clients on everybody’s desktop.
But it turns out that those benefits accrue to something called “network computing,” of which VDI could be considered a subset, not VDI itself.
What’s wrong with VDI?
Well, first of all, the hardware involved costs 40-80% more than just replacing a desktop with a new one. The economics of PCs are so well trodden that it’s difficult for any other hardware configuration to deliver as much value at the same price.
Then, there’s the question of service level. When one desktop goes down, a single worker (two, if you count the deskside IT help) becomes nonproductive. But if there’s a glitch at the server end, 10,000 workers could be down for an unknown period of time. Building in redundancy might be a solution, but then VDI costs 100-200% more than the simple PC alternative.
ClearCube, a Texas-based specialty PC hardware and VDI software OEM, has made a decent niche business out of a form of VDI that involves a mix of PC client types that live in a rack in the datacenter. On the desktop sits a keyboard, mouse, monitor, and a “puck” (a small thin client that communicates over an IP link with the datacenter). ClearCube’s connection broker assigns requests from the desktop to either a particular rack-based client, maybe with a video accelerator operating between the two for workstation-class service, or to the next client available.
Although ClearCube has addressed many of the complexity issues that plague VDI, this elegance doesn’t come cheaply. ClearCube’s customers (e.g., the military) are more than averagely price insensitive.
And if the user doesn’t have one of those neat video-accelerator channels, he or she often must contend with irritating latency and a lack of responsiveness. A movement of the mouse is followed agonizing seconds later by a change in cursor position.
VDI also has a problematic value proposition: if the IT manager spends a big pile of money today, he or she will see an unquantifiable savings in operating expense in the unknown future.
Meanwhile, there’s a lot of work the organization must do to prepare for VDI. For example, it must take its own image of, say, Microsoft Office and publish it as a unique application.
But people need and want networked computing. And the answer may lie with the cloud, with solutions coming from Web-based applications rather than published applications or terminal services.
Web interfaces are friendlier, which fits well with the overall move toward the consumerization of IT. And everybody has a browser: PCs, phones, tablets, even the Nintendo Wii and smart TVs. By making the browser the thin client, the cloud service can address the entire world’s devices.
Using application virtualization, IT managers can take an application and wrap it in an envelope, which can talk to any OS (e.g., Google‘s Android, Apple‘s OS X, Windows Vista, Windows XP, any service pack). Most IT shops manage a corporate image, a string of bits containing a tested compilation of OS, drivers, and applications that, when laid down on a fresh hard drive, just runs. The problem is, every time Microsoft releases a service pack, the image has to be tested all over again, a process that often takes weeks or months. In the Web-based envelope model, all that has to be updated and tested is the envelope itself, a solution quite pleasing to IT, which would love to get off the validation treadmill.
This type of application streaming is available from Citrix as XenApp, from Microsoft as App-V, and VMware as ThinApp (formerly Thinstall — I guess they didn’t like the implications of the word “stall”).
These application virtualization packages are a good way to pull in regular old PCs as well as VDI.
Unfortunately, complexity doesn’t go away. It just gets shifted elsewhere. Images get bigger. There’s more to maintain, and then there’s resource contention on the network or at the server level.
So, we’re back to browsers, which ultimately offer the best solution. Browsers will get smarter and more ubiquitous and will manage and take advantage of inexpensive local resources on the endpoint (i.e., PC, phone, game console, washing machine). HTML5 will make Web applications easier and more powerful.
With the exception of expensive, integrated solutions like ClearCube, which will remain quite limited, VDI is dead meat. Many large companies that started pilots a few years ago have halted deployments. They don’t have enough bandwidth or servers. The environment doesn’t scale.
© 2012 Endpoint Technologies Associates, Inc. All rights reserved.