From Server Farms to Consolidation…

Mainframes were once expected to be replaced by smaller, more nimble commodity servers. But as commodity servers proliferated, their inefficiencies became clear. Ten to 20% CPU utilization rates were common, and data centers filled with barely-used disk space.

Statistic: According to Forrester Research, Windows servers typically run at only 8 to12% of their full capacity, and UNIX servers at only 25 to 30%.

Still, commodity servers are low cost and easy to replace; initial hardware expense is not a major consideration when deciding to install a new server. However, the initial hardware expense is dwarfed by associated costs, like administration costs, software licenses, power, cooling, and floor space.

The emphasis has now shifted to server consolidation to reduce overhead and contain costs. Server virtualization technology has made server consolidation easier to accomplish, helping to fulfill the promise of an inexpensive server-based architecture. Unfortunately, monitoring, managing and capacity planning for virtual resources presents a unique series of challenges.

…to Virtualization

In virtualized server environments a layer of abstraction separates what the software sees from the actual hardware. This layer allows hardware to be treated as a pooled resource shared by several guests.

The diagram above shows two guests running on a physical host. One guest is running Red Hat Linux, the other is running Windows 2000.

Tip: Look for tools that aid in the seamless management of complex, heterogeneous, virtualized environments for a competitive edge.

The ability to understand these four aspects of virtualized servers is critical to ensuring optimal performance:

1. The container as a whole, including overall CPU and memory usage, so underutilized resources and those nearing capacity can be identified

2. The service console

3. Individual guests, to be sure sufficient resources have been allocated to meet performance requirements

4. Guest interaction, to be sure the applications running on each guest are compatible

Monitoring, managing and capacity planning for server virtualization presents unique challenges. For example:

1. At what level of activity does a guest exceed the capacity of its host system?

2. When will a guest exceed a limit for a resource?

3. When will a guest exceed its virtual processor configuration?

4. How much of my physical host is being consumed by each guest?

5. If my application is using 50% of a CPU, what does that mean? What is 50% of a virtual CPU?

Good performance tools that can answer such questions are not yet common, but it is worth the effort to find them before implementing a server virtualization strategy.

In summary, though more complex than traditional server environments, virtualization paves the way for IT infrastructure optimization and cost reduction. Improved manageability, decreased floor space requirements, and increased flexibility are also positive side effects, if not impetuses for going virtual. However, the critical need for careful performance management and capacity planning practices must be addressed to realize these benefits.

Ron Potter is IT Best Practices Manager at TeamQuest, 641-357-2700 x2238; email: ronald.potter@teamquest.com; web: www.teamquest.com.