In my story about managing virtualization, I recommend how IT managers can tame existing server sprawl. This sprawl can be avoided altogether at companies just beginning or planning a server virtualization if the following steps are taken:
Know your existing physical network. IT managers must leverage existing knowledge of the physical systems along with the logical network layout of those systems. Virtual systems must exist on physical hosts, and these physical hosts provide clues about the most effective architecture for implementing virtual systems.
Survey workloads. Next, survey the workloads that will be virtualized. Make sure that business-critical applications are certified to operate in (and are supported in) a virtualized environment. If not, find out when the vendor intends to provide such certification. If you can't determine that, it is likely not advisable to virtualize the application.
Chart apps. For applications that are supported on a virtual platform, make a chart of the performance, security, patch, update, backup and recovery characteristics of the applications. For x86 environments that are still relative newcomers to the virtualization party--as compared to mainframe systems that have been managing virtual workloads for many years--it is advisable to put similar workloads on the same physical host.
Baseline costs. Everything I have recommended up until this point will be beneficial even if you don't go forward with a virtualization plan. If you are going to move forward, now is the time to baseline current operating costs and get at least a basic understanding of how costs will be charged to users of virtualized infrastructure.
Get ready to start charging. While researching my managing virtualization story, it was clear to me that one thing that is lagging in virtualized environments is a good chargeback mechanism. In the era of running workloads on individual physical systems, chargeback was fairly straightforward. In a virtualization era, chargeback is going to take on a whole new level of importance. Groups that may have been willing to pay for servers reserved for their exclusive use will likely want a different price structure when sharing the physical resource with different parts of the organization.
Look to SAAS (software as a service) multitenancy for chargeback models. The idea of multitenancy that is all the rage in the SAAS world has some relevance for chargeback. IT managers should look into usage measurement tools so that they can provide business users with fairly accurate figures of what it costs to get into the virtualized infrastructure. As energy and cooling costs escalate and rack space becomes more scarce, chargeback billing will become a core component of managing a virtual server infrastructure.
Build a lab based on your real network. Finally, build a lab using the equipment and workload present in your organization based on organizational surveys conducted by the IT department. Assess the impact of using various virtualization platforms on CPU, RAM, I/O and storage. In particular, if your organization uses multiple virtualization platforms, ensure that you get a management tool that can stretch across these platforms to provide the operational details needed to maintain control of the infrastructure. Baseline the costs of providing virtual infrastructure using the lab to get a starting price that business groups will pay to use the new virtualized environment.