Maximum Virtualization: Addressing the Confidence Obstacle

By Beth Ruck
published: Wednesday, October 08 2008

I spend a lot of time these days talking to customers, project managers and systems engineers about data center management challenges - and much of the conversation revolves around managing virtualization.  From the system engineering standpoint, virtualization is a no-brainer for many types of applications.  Who wouldn't want to increase server utilization, condense the data center footprint and increase agility in deploying new servers, when improved efficiency and cost savings is so often the result?

So I was surprised when I started hearing that the greatest obstacle to widespread adoption of server virtualization was not any particular technical issue or skills-related issue, but an organizational one:  line-of-business stakeholders would not allow it.  As attractive as the financial incentive might be, many were - and are - afraid that virtualization would put their applications at risk.

Fair enough; their revenue streams, staff productivity and job security probably depend on the performance of those applications.  It is not easy either technically or organizationally to migrate to a shared resource model when, for so many years, the answer to the fear of poor performance was to buy more bandwidth or processing power than was really needed.  Let's face it - many organizations are still more comfortable with that model, and lack the staff resources and management tools to easily prioritize a more methodical approach to capacity planning.

So how can proponents of virtualization deal with the lack of confidence among business managers?  A surprising number of people I talked to at VMworld last year were taking a guerrilla approach - virtualizing covertly and asking for forgiveness later, hoping to demonstrate after the fact that performance fears were unfounded.  While that may be the most expedient way of handling it, it is potentially a risky one. After all, driving up server utilization levels does tend to put application performance at greater risk and virtualization can significantly increase the impact of a single instance of system failure.  It can make it harder to identify the physical root of performance degradation unless you've got a pristine process for keeping track of where each virtual machine (VM) resides - or management tools that automate the inventory task.  And finally, shuffling VMs from one physical host to another can suddenly cause network congestion - and pose risk to many services - if the larger network picture is not properly taken into account.

A better approach to addressing this lack of stakeholder confidence in virtualization is to implement service quality reporting.  The call for reporting and accountability on end-to-end service quality is being made in most organizations anyway, with or without virtualization.  While nothing can magically make organizational challenges go away, champions of virtualization should seize and help accelerate this trend.  Demonstrating that you have a real, proactive handle on system behavior and performance requirements in the context of the whole service infrastructure would go a long way toward assuring the line-of-business teams that you have the visibility needed to preempt the problems that can arise when implementing a virtualized architecture.