Network World recently published an article around the management challenges of Virtualization where they surveyed more than 400 IT professionals on their virtualization strategy and how they manage their virtualized environments.

The outcomes of the survey are not surprising.   Essentially what the article is saying is that those organizations acknowledge the benefits that virtualization brings, including cost savings, agility and so on, but they lack the tools required to manage such environments.

That’s also what I’m hearing when I speak to IT organizations with highly virtualized environments.   Traditional infrastructure management approaches simply do not work in virtualized environments.   The standard infrastructure counters that are measured are no longer useful to track the performance of a virtual environment.

For example, how representative are traditional metrics like CPU, IO, Swap Space, Memory usage and so on, when the environment is constantly changing?  How can you compare these values to previous values when the underlying resources are being changed to meet demand, and new virtual machines are being created/removed to meet application loads?    Is high/low resource usage even a problem?

Not only that, but many organizations used to rely on experienced operations folks who understood what infrastructure supported what applications/business processes.    In a virtualized environment, relying on this sort of knowledge is much more difficult as the environment is changing so fast.

Realistically the only way to manage such environments is to understand what the business needs of each key application and how well the applications are meeting that need – essentially relegating infrastructure management tools from being at the forefront of problem identification to more of a supplementary information role, with application and end user experience monitoring solutions providing the core focus of what is good performance and what is bad.

This represents a fundamental shift for many organizations who have built their entire incident management processes around reacting to anomalies or alerts from infrastructure rather than monitoring the application.    These organizations are going to have to come up with a good strategy for monitoring the application and the end user experience to cope in a virtualized environment.

The article did not talk at all about virtual machine sprawl – which is a huge issue for highly virtualized environments.   This was a big surprise as I would have expected it to get a mention.   Virtualization is great that its so easy to stand up a virtual machine whenever its needed – especially the ability to be agile and respond to business needs.   But who manages these and decommissions them when no longer needed.   Many organizations are really struggling with this process.