May 21, 2009, 3:55 PM — Automation tools are important in preventing virtual server sprawl, but IT managers have to set strict policies and enforce them to avoid unnecessary proliferation of virtual machines.
That’s the message from a panel of virtualization management vendors that discussed sprawl at Interop Las Vegas this week.
While the vendors sang the praises of their own products, they said human processes may play an even more important role than technology in controlling VM sprawl.
“We found sprawl was more of a behavioral thing than a technology barrier,” said John Suit, CTO and principal founder of Fortisphere. “It has a lot to do with the expectations of application owners.”
The most successful IT shops set clear expectations and don’t divert from them, said Jay Litkey, founder and CEO of Embotics.
“We have enterprises with thousands of virtual machines, and they don’t have sprawl,” Litkey said. “They’re very disciplined, they do a lot of planning, and they have a lot of processes.”
The biggest hurdle may be politics, which George Pradel called the “eighth layer” of the network stack. Pradel, Vizioncore’s director of strategic alliances, said enterprises have to set policies that virtual machines will be shut off or decommissioned if they are not in use for a certain amount of time. Without planning, virtual server sprawl will get out of control, he said.
“If you are investing in some type of virtualization and you haven’t [experienced sprawl] yet, you will,” he said.
Interop attendees are embracing virtualization in a big way. An on-site survey of 120 network engineers and IT managers and executives found that 55% have virtualized mission-critical servers. But virtualization caused more problems than benefits for greater than half of the respondents. The survey was conducted by Network Instruments.
Sprawl is one of the most common problems related to virtualization, said Anne Skamarock, who moderated the sprawl panel and is research director for Focus, an IT consulting firm. Nonfunctioning virtual machines still take up memory and space, and must be controlled, she noted.
While virtualization can simplify life in many aspects, including disaster recovery, panelists said it introduces many new complications. With virtualization “there are two layers of infrastructure,” said Dave Malcolm, CTO of Surgient. “You have to be able to keep track of the virtual machines’ relationships to physical infrastructure.” Maintaining performance through these layers of abstraction is another issue.
“If virtualization wasn’t hard to manage or [didn't] have its own complexities, we wouldn’t be in business,” Litkey said. “In many respects, we’re still learning as we go. Companies are still running into new challenges.”
The four vendors on the panel make tools that integrate with VMware and other hypervisors, automating the management of virtual machines and enabling features such as self-provisioning, automatic discovery of rogue VMs, and visibility into usage and capacity.
If you have a small virtualization deployment, a spreadsheet might be a sufficient tool for keeping track and managing virtual machines. But once the deployment grows large enough, a third-party management tool may be necessary, panelists said.
“When you’re over 100 VMs, you probably ought to be looking at an automation solution,” Malcolm said.