April 16, 2008

By Alan Radding

Virtualization has played a part in the enterprise for decades, but only recently has it taken on new urgency as organizations wrestle with information technology (IT) infrastructure sprawl, skyrocketing energy bills, cooling challenges, and data center space constraints. Initially a mainframe phenomenon, virtualization has emerged as the preferred strategy of organizations everywhere, leading Yankee Group, a Boston-based technology research and consulting firm, to declare virtualization the future of IT infrastructure.

Today many organizations are still feeling their way toward virtualization and learning many lessons along the way. As an IT strategy, virtualization is more than simply a product to buy. "There are a number of different approaches to virtualization. Before you get started, you need to think about the problem you are trying to solve," says Gordon Haff, principal IT advisor at Illuminata Inc., a research firm based in Nashua, N.H. In addition, organizations embarking on virtualization need to think about managing the virtualized environment, high availability and disaster recovery, hardware sizing, network fabric planning, and more.

The problems most organizations are currently trying to solve with virtualization include the high cost and constrained availability of energy, the need to reduce data center heat to lower air conditioning costs, and overcrowding in the data center. The solution generally comes down to server, storage, and network infrastructure consolidation.

"We had a lot of processes on traditional 1u and 2u servers, and we were running into a heat issue in the data center," says Rick Chin, senior vice president for information technology at Pinnacle Financial Corp., an Orlando, Fla., independently owned direct mortgage lender.

Chin turned to virtualization, initially running about a dozen servers as virtual machines under VMware on a Dell blade server connected to a Cisco Catalyst 6513 Switch. "We were able to bring down the temperature about 15 degrees, which reduced the load on our AC," he reports.

According to Chin, Pinnacle did several things to drop temperature in the data center, including switching to the blade server and turning off many of the 2U servers. This dropped the temperature 7 to 8 degrees, and other things dropped the temperature the other 7 to 8 degrees.

Lessons Learned

Virtualization is a key to infrastructure consolidation. "The industry is coming full circle as it recentralizes the infrastructure. With the value of virtualization compounded by the number of resources being virtualized, it pays to recentralize and virtualize all you can," says Doug Gourlay, Senior Director of Marketing and Product Management for Cisco.

Another problem organizations are solving with virtualization is infrastructure availability. "With server virtualization, for example, you can use mobility instead of taking down servers," Haff says. By mobility, Haff means the ease of moving virtual servers between physical servers in a cluster. When a physical server needs maintenance, the virtual servers are easily moved to another physical server without interrupting service to the users. Through the use of a storage area network (SAN), the administrators do not even have to redirect the storage. Advanced virtualization software will even automatically move virtual servers to another server in the cluster in the event of a failure.

"The faster you can move a virtual machine, the easier it is to solve business continuity problems," Gourlay says. Virtual machines are very easy to move over a network; everything is encapsulated in a single file. And when the data is on the SAN, "you can move the server without even moving the data store," he adds.

Simple as virtualization sounds, there are some pitfalls to avoid. First, you need to plan capacity and size the hardware correctly. "You want to start with a baseline assessment," Gourlay says. This assessment looks at the existing infrastructure and the utilization of each component individually. Understanding utilization is the key to successful virtualization.

"You want to take a holistic view of the infrastructure," he continues. Too often organizations just look at server utilization. In a holistic approach, you look at storage and network as well as servers. "You need all three to process the workload," Gourlay notes.

Although you can take a lightly utilized server, add virtualization software, and start creating virtual servers, most experts suggest starting with new hardware. Virtualization carves up and manages the physical resources so that they can be shared by multiple virtual machines. Starting with more processing power, memory, and bandwidth provides a distinct advantage.

"The next time you have a planned infrastructure build out is the time to deliver an end-to-end virtualization approach," Gourlay says. The new servers, for example, will likely have faster processors and more memory. Similarly, new switches will deliver greater bandwidth and capacity, which will come in handy as more virtual servers vie for server and bandwidth resources.

Deploy virtualization on existing servers when you only want to test virtualization and determine the operational requirements. In that case, add virtualization software to the server and experiment with virtualization. You will gain virtualization skill and experience at very low risk and be ready to move to enterprise-class virtualization with confidence.

Finally, you will need to consider management. "In some ways management will be harder; in other cases it will be easier," Haff says.

Management becomes harder, because your physical servers will have to be managed for maximum availability and reliability. You are putting many virtual applications on one physical server-all your eggs, in effect, are in one basket-so you cannot allow it to fail. That means redundant everything, failover clusters, predictive analysis to avert failures, and more.

What is not more difficult is managing those many virtual servers. That gets easier, because server virtualization products typically include effective administrative consoles. Tools for managing the rest of the virtualized data center are in a nascent stage, though.

"The tools will be there when the customer needs them," Gourlay says. Cisco, for example, offers the Cisco VFrame Data Center, and orchestration platform that enables an organization to manage its compute, networking, and storage resources as a set of virtualized services. More tools will be coming.

And do not forget the people. Along with the right tools, another lesson learned is to have a technician on your team dedicated to learning and managing virtualization. Pinnacle's Chin says having a particularly good technician on his team who embraced the challenge was another key plus. "It's not difficult to set up virtualization, but it's always helpful to have an enthusiastic technician," he says. "And it doesn't require much maintenance once you have the virtual machines running."

Virtualization Lessons Learned

• Plan capacity and size the hardware correctly
• Take a holistic view of the infrastructure, including storage, network, and servers
• Start with more processing power, memory, and bandwidth
• Experiment with virtualization by adding virtualization software to an existing server, virtual network services, and SAN virtualization
• Manage for maximum availability and reliability; that means redundant everything
• Dedicate one team member to learning and managing virtualization

Alan Radding is a freelance writer in Newton, MA.