The rapid adoption of server virtualization at many global enterprises is leading to the growing issue of “virtual stall,” according to Andi Mann, former Enterprise Management Associate analyst turned CA Inc. product marketing executive.
This issue arises after organizations have grabbed the initial cost savings from consolidating their physical machines and are now running into other challenges such as a growing production environment, more pressure on staff and skills, and increasing line of business visibility.
Speaking at a session from VMworld 2010 in San Francisco, Mann cited CA data that showed 30 per cent of servers are virtualized at the average global enterprise. He added that this benchmark is also where most companies first encounter this new virtualization hump.
While low-risk test/dev servers and Web servers might have been virtualized with tremendous results, Mann said, many mature virtualization shops are having difficulty getting buy-in for existing servers that house critical business apps, customer-facing apps, and other composite apps.
Knowing what you have
To get over “the hump,” Mann said organizations must first gain more visibility into their environment. “You can’t manage what you can’t see,” he said.
For Mann, visibility starts with being able to track and record where your VM hosts and guests are running and how they interact with each other. This is especially important if a company has any desire to move to a hybrid, private/public cloud environment.
The audit trail should also extend to physical systems and apps, not just the VMs, he added.
With this information, IT can create detailed inventory lists, recover unused machines and licences, and ultimately, deliver this data to business owners to gain additional virtualization support.
Jon Schulman, a senior systems engineer at McLean, Va.-based SAIC Inc., said his company is working as a contractor to a joint U.S. Army/Navy data centre facility. He is using VMware products, such as vConfiguration Management to help manage his environment.
For Schulman, the fact that the software treats physical and virtual infrastructure the same way is critical for his asset control needs.
“VMware is just another server brand to us,” he said. “That’s how we want to think about it.”
Controlling what you have
After getting a handle on what you have, Mann said, organizations can go beyond monitoring and start taking action. He advised organizations to look for tools that can take control of replication, migration and continuity of VMs moving not just from physical to virtual infrastructure, but also between virtual servers.
It is also important to centralize control as much as possible, he said. This is especially useful if your organization has dozens of different groups with varying levels of expertise.
“Centralizing the management and integrating it eases the skills requirements and the burden on staff,” Mann said.
Making good on guarantees
To get app owners to give up physical hardware and keep buying in to more virtualization, IT needs to assure them they can deliver services.
“Performance is great, but performance isn’t necessarily good or bad,” Mann said. “You don’t want to make sure you’re delivering performance. You want to deliver assurances.”
Mann said tracking CPU usage is a pretty pointless measure because it means nothing in terms of the performance of the apps themselves.
For instance, tracking and monitoring what your VMs are doing how they’re interacting with each other can give IT the information it needs to assure the business that they will be compliant with regulatory laws, he said.
A key feature to look for in a VM management tool, he said, is the ability to do auditing in real-time, as opposed to just tracking historical data.
Set it and forget it
The last piece of the puzzle, Mann said, is the ability to automate all of these processes.
“If you’ve got enterprise architects provisioning virtual servers, you’re wasting your time,” he said. “Think about how you can use repeatable provisioning with reusable VM templates.”
Scheduled deprovisioning of VMs is a vital weapon in the battle of VM sprawl, he said. Building the same repeatable process for creating and patching VMs will make the overall system more efficient and help you reduce errors.
Mann’s advice to getting started with automation is to standardize one or two good processes.
“If you standardize the two steps in a 50 step process, you’re getting there,” he said. “A thousand mile journey starts with a footstep.”