Avoiding virtualisation’s dark side

Uncontrolled virtual machine sprawl can undermine the business case for virtualisation, says Warren Olivier of software vendor Veeam.

January 28, 2013

Uncontrolled virtual machine sprawl can undermine the business case for virtualisation, says Warren Olivier of software vendor Veeam – and IT managers need to deploy new tools to manage much more complex environments.

Olivier identifies three major issues in managing virtual environments: Sprawl, rightsizing and monitoring.

“Virtual machine sprawl happens precisely because these machines are so easy to create and move around,” he says. “If your development team needs five machines to test their code on, you can deliver in 10 minutes instead of waiting weeks to go through a procurement process. But do you remember to delete those five machines again when they’re no longer needed?”

“People tend to go crazy,” says Olivier. “Then you’re using up more disk space, more memory, more CPU cycles, more power… in the worst cases, you could end up with exactly the same costs as before you virtualised, with extra licence fees added on top.”

The problem is worsened where virtual machines are not rightsized, adds Olivier. “Most companies have a VM template they roll out as standard: 2 CPUs, 8Gb memory, so much storage space – but does every machine really need that much? In many cases 4Gb of memory is quite enough. If every machine you create is over-specced, your consolidation ratio drops – which is a problem because that’s what the business case is based on.”

Resource wastage can threaten the promised benefits of virtualising. In the same light, Virtual Machines that are ‘starving resource’ are not efficient and could be running poorly. With right sizing reports, customers can run meaner and leaner and reach the ROI quicker.

Typically, says Olivier, his clients based their business case on a consolidation ratio of eight virtual machines to one physical server – but are running at only six. “Really what you want is 15 or more, based on underlying hardware specs of course,” he says; “but you can only size your machines to that level of precision if you know the history of each virtual machines resource utilisation and look back and project its needed resources based on trending.”

Which introduces the final problem: Monitoring and keeping track of virtual environments that are constantly changing. “There are lots of new metrics that are unique to virtualisation,” says Olivier. “Machines are always moving within clusters to optimise their resource use – that’s the whole point. But it also means that all your documentation is out of date hours after you’ve produced it. It’s simply not possible to monitor a virtual environment manually.”

As a result, says Olivier, many companies are starting to monitor their virtual environments – but are not always choosing the right tools. “You can’t put agent software designed to monitor a physical machine into a virtual machine,” he says. “Virtualisation is all about tricking the operating system into thinking it has access to more resources than it really does. So when the agent reports that you’re at 100% memory utilisation, you’re not getting accurate information. Virtualisation allows you to over-commit on memory, therefore you need to monitor your virtual machines from the outside, not the inside.

Olivier says Veeam Management Pack for Microsoft System Center Operations Manager and Veeam ONE, provide a comprehensive virtual environment management, reporting and monitoring solution.

“If you manage it properly, the benefits of virtualisation far outweigh the dark side,” he says. “The fact that virtual environments are so flexible is their greatest strength – but things can get out of hand really quickly. Fortunately, if you use the right tools it’s easy to stay on top of things.”