As important as it is for administrators to make effective use of the IT budget, careful budgeting is even more important if you're responsible for managing a virtualization infrastructure. After all, virtualization environments are larger and more complex than ever before, and cloud layers are also being thrown into the mix, further increasing the cost and complexity.
However, the IT budget typically doesn't keep pace with infrastructure growth, so you must employ creative virtualization cost-savings methods.
Review your licenses
One of the most effective virtualization cost-savings techniques is looking for wasted licenses. There are a few different ways that licenses might be wasted.
Perhaps the most common source of wasted licenses is subscriptions that have been forgotten. This happens when an organization subscribes to a cloud-based service and then discontinues use of the service halfway through the subscription. By the time that the subscription is due for renewal, it has been all but forgotten. The organization might not even realize that it has been renewing a subscription that it's no longer using.
A second way that licenses are wasted is in the form of overlapping tools. Native hypervisor management tools such as vCenter Server or System Center Virtual Machine Manager are great, but larger organizations almost always depend on third-party tools that provide capabilities that extend well beyond what the native tools are capable of.
Over time, third-party tools inevitably become more feature rich as they progress from one version to the next. As such, an organization might eventually discover that it has multiple tools with nearly identical capabilities. In such situations, the organization might be able to save a considerable amount of money by re-evaluating its choice of tools and eliminating any redundancy.
One more way an organization might be able to achieve virtualization cost savings is by examining the way it licenses host servers. Let's suppose for a moment that a small organization initially decides that it only needs a few virtual servers. It decides to purchase two Windows Server Standard Edition licenses, which allows for a total of four OS instances. Later on, the organization decides to add another VM, so they add one more Windows Server Standard Edition license to the stack.
Even though Windows Server Standard Edition licenses are far less expensive than Windows Server Datacenter Edition licenses, a Datacenter Edition license allows you to run an unlimited number of OS instances on a licensed host. As an organization adds Standard Edition licenses, it will eventually reach a point at which it is more cost-effective to simply purchase a single Datacenter License than to keep paying for all of those Standard Edition licenses.
Increase VM density
Another very effective virtualization cost-savings technique is to look for ways of increasing a host's VM density. Doing so enables the organization to avoid having to spend money on new servers to accommodate new workloads. It also helps reduce licensing costs, as the organization won't have to worry about licensing a new server.
Although it isn't always possible to increase a host's VM density, there are a few things that you might be able to do. First, check to see if the server can accommodate additional physical memory. Memory is almost always one of the factors that limit the total number of VMs that can be run on a host. You should also consider enabling dynamic memory if it isn't already being used. Dynamic memory enables a host's existing memory to be used more efficiently, thereby potentially enabling the VM's density to be increased.
Another possibility is to invest in better performing storage. Storage IOPS is a very common limiting factor when it comes to virtualization deployments.
As you work to improve the overall VM density, just be sure not to use all of your host's available capacity. Most production virtualization deployments are clustered, and so you must reserve enough capacity to enable the cluster to absorb a host's highly available workloads in the event of a node failure.