Marcia Kaufman, COO and Principal Analyst
Sponsored by IBM
As IT organizations work to increase the flexibility, efficiency, and cost effectiveness of their data centers, many have focused on server virtualization. These companies are also leveraging server virtualization as an important starting point for their journey to adopt cloud computing. However, server virtualization alone is not sufficient to help companies realize the full compliment of cloud computing benefits. In essence, to achieve a wellmanaged data center requires a more application centric approach. One way to ensure that the end user is supported as part of the virtualization strategy is to make the resource allocation process application aware.
Why is this issue so important? The use of applications is not static. It is common that during certain business cycles, the demands on a particular application will increase and therefore require additional unanticipated computing resources. When this happens in a heavily virtualized environment, the virtual machines that are supporting the application may not be able to handle the changing workload.
Companies need a way to prioritize applications not only based on infrastructure requirements and policies, but with an understanding of the business context of the applications in question. In order for companies to ensure that the most business critical applications are given top priority, its service polices must reflect the relative business impact of each application. Server virtualization allows the addition of entirely new virtual machines to an application, but this takes time, and the spike could be gone before the new resources are ready. Furthermore, the action of adding a new virtual machine is only useful if the system knows when to use it. Server virtualization products can look at system metrics such as memory and CPU usage, but this gives only a very high level view of what’s really going on, and completely ignores any difference in the priority of different workloads in periods of resource contention. As such, with server virtualization alone these additional applications may suffer poor performance even if they have a higher priority to the business.