Virtualization Quick Hits

• By Stuart J. Johnston
• April 24, 2006

“Virtualization” is increasingly a word bantered about in corporate boardrooms. Everyone from the CIO to the CTO to the CFO – even the CEO, in some cases – has heard of vast benefits and, better yet, decreased IT costs, related to going virtual.

No. Not virtualization as in moving all of your workers off the company premises – a so-called virtual workforce – but rather virtualizing the computer systems they use. As the computing world gets more complex, something has to help simplify management and maintenance of those systems, and help keep ongoing costs under control.

But there are a lot of different forms and flavors of virtualization and a lot to know, even before you start on virtualizing that first major production system.

For many, the question is How to get a toe in to test the waters without getting scalded on the first attempt.

In recent conversations with vendors and analysts, a variety of fairly small, quick to implement, and fast to provide return on investment projects emerged -- ones that can not only help IT staffers look like heroes to management, but can also yield valuable experience as well as immediate benefits.

Although there are several options for virtualizing users’ desktops, either on their own local hardware or on a remote server, this article focuses on the use of virtualization on the server side. These are typically cases where the server hardware runs on a base operating system – such as Windows Server 2003 -- or on a so-called “hypervisor” – and multiple guest operating systems, each running their own applications, run on top of that platform.

One point to think about is not doing the most mission critical applications first.

“Today, people are trying [virtualization] on their non-critical apps,” says Gary Phillips, president and CEO of virtualization software vendor Marathon Technologies. Experience gained in early projects, whether successes or failures, can be valuable later when converting line of business applications to virtualized environments. Littleton, Mass.-based Marathon Sells low-cost fault tolerant and high availability solutions, that run on top of Windows Server 2003, built on its own virtualization technologies.

But where to start?

Jim Ni, group product manager for Microsoft Virtual Server has a list of areas where he feels IT can get the most, and quickest, bang for the buck – server consolidation, and testing, and what he refers to as “legacy application rehosting, as well as development and testing, and interestingly disaster recovery.”

Server consolidation is at the top of the list for a good reason.

“What [customers] find is that when they go into a production project, they find they can get more [virtual] servers on a smaller number of boxes,” says Al Gillen research director for operating systems at researcher IDC.

In fact, server consolidation has been one of the most popular, if not the most popular, uses of virtualization to date. Both market leader VMware and second-place Microsoft have products – VMware Server and Microsoft’s Virtual Server 2005 -- to run multiple guest operating systems on a single hardware server.

“With pure server consolidation to get just a much more efficient [use of computing resources], we see a three to six-month ROI,” says Dan Chu, senior director of developer and ISV products at VMware in Palo Alto, Calif. Much of that comes in savings on server support costs, he adds.

There’s also a bit of incentive in the fact that both VMware and Microsoft have made their products free, say analysts. “[Customers] can get a sense of what the value proposition would be [and] as far as try-before-you-buy, you couldn’t wish for a better scenario,” says IDC’s Gillen.

Legacy application rehosting is a similar area for quick implementation and fast payback from virtualization pilots. IT shops often have older applications that need to run on a specific version of an out-of-date operating system – say Windows NT 4.

Maintaining the legacy application on obsolete hardware may no longer be cost-effective or secure, but the app is still needed. System-level virtualization would let those shops run NT in a virtual machine, so that the application would appear to be running on the old hardware, but would actually be running on the latest hardware and software (underneath it all).

Running server apps in VMs can also be useful for increasing production flow through or for overall system load balancing. Because they are contained in VMs, a virtualized application is easier to administer than an application that is physically scattered across a server farm. “If I had a payroll app that spikes every two weeks, [because of virtualization] I could migrate other applications to other servers [for that time period],” adds VMware’s Chu.

Another area cited by both Chu and Ni as a quick winner: virtualizing system and application development and testing environments. IT shops often have to dedicate entire servers – even if that turns out to be hardware overkill – to these processes. In these cases, a system-level crash caused during a debugging operation could easily take down the whole server. Therefore, these environments often need to be isolated from any production systems and applications, usually meaning separate servers for each.

By running the development and test environment in a VM, however, when there’s a crash, only the VM goes down with no effect on any other VMs running on that server. One application crashing won’t take down others around it. Recovery is faster and more than one VM can be run on a single server, making for better manageability as well as potentially lowered hardware costs.

Interestingly, both VMware’s Chu and Microsoft’s Ni also see quick projects with fast payoff in the area of disaster recovery.

While he agrees about the other three areas, Rob Enderle, principal analyst at consultancy the Enderle Group, has his doubts about virtualization as a quick hit project for disaster recovery. “That’s a stretch . . . it just doesn’t seem to be that practical,” he says.

Still, Chu says it’s a popular area for customers. “63 percent of our 20,000 plus enterprise customers are using VMware [for] disaster recovery,” adds Chu.

Indeed, Ni backs up Chu’s statement. “[The recovery server] could be hosted on completely different hardware – [perhaps] older hardware [than the primary server] with different production applications,” Ni says.

Get the latest download of VMware server here.

Find the latest download of Virtual Server 2005 R2 here.