Managing Server Virtualisation:

From NetworkComputing Magazine Vol 17 Issue 01 - January/February 2008

Beth Ruck of InfoVista explains some of the considerations for managing virtualised servers

Most companies have jumped into server virtualisation with both feet. Though most are a long way from achieving ultimate system utilisation targets, initial phases of virtualisation have enabled IT teams to make important advances. Firstly, they're familiar with the technologies that are now part of the bedrock of data centre architecture. Secondly, most have proven to managers and CIOs the value of the technology and have thus overcome significant organisational obstacles.
Attention, rightfully, is beginning to focus on the challenge of how to manage the virtualised environment, which is driven from two perspectives.

System-Focused View
The system administrator must have a close eye on real-time system capacity issues, not just box-by-box, but across groups of servers. As physical resource utilisation increases, the risk to server health increases, because there is less spare capacity available to absorb spikes in user demand. Without the adequate tools in place to accurately inform the resource allocation process, virtualisation rates will level off and the business will not allow application performance to be at risk.
Although most companies have a variety of vendor-specific management tools - some will say too many - they lack visibility across the whole server farm. Creating global visibility to see where resources can, and need to be reallocated, requires sophisticated ability to collect, normalise and weed out the truly relevant data from multiple servers. This is no small task, especially in a typically heterogeneous environment.

Service-Oriented View
However, virtualisation ripples more broadly. It accelerates the second driver - the call for service-oriented management of the data centre, usually coming from a CIO or senior IT perspective. What is really meant by service-oriented? IT is judged not by server or device uptime, but by the functioning of applications and the services they comprise. In the virtualised world, where application performance is no longer neatly tied to the health of statically-assigned resources, we need a dynamic picture of exactly what resources are linked to which services, at any given time.
Service-to-resource linkage serves problem-solving from two angles. When users complain that a service is slow, we need a view of the entire supporting infrastructure including servers, LAN and WAN links, switches, firewalls and so on, to see where along the delivery chain there are red lights or warning signs. From the other direction, when an element malfunctions, we need to know which services are impacted to prioritise corrective action.

Which Way Forward?
Technology buyers face a dilemma about how to invest in virtualisation management. Dozens of emerging products provide granular information about virtual machines. Some also automate tasks associated with resource allocation and other "hot" concerns. But, just as most environments are saddled with many platform-specific system tools, adding disjointed virtualisation management products just adds to data centre complexity - and therefore adds to cost as well. Companies that address virtualisation management from a system-focused standpoint should select a foundation that provides both a virtual and physical view across the server farm; and one that can integrate with niche products if needed.
But ideally, this should also be a service management platform that will accommodate the server, network and storage aspects of virtualisation, as it becomes integral throughout the infrastructure. Service management, because it inherently extends across silos, pushes the organisational envelope. It streamlines management by providing a single source of information shared between IT domains.
Virtualisation is therefore more than a revolution within the server domain itself. By addressing the urgent need to manage it, the data centre manager can champion the breakdown between technology silos that is the Holy Grail for most CIOs. NC