When it comes to server virtualization, there seems to be an endless debate over whether it is better to virtualize all of an organization's servers or to leave some running on physical hardware. While there is no definitive answer, there are a number of points to consider when determining whether a fully virtualized infrastructure is a good idea for you.
A virtualized infrastructure provides unmatched flexibility
Although hardware consolidation is often cited as the top argument in favor of server virtualization, those who recommend virtualizing all of an organization's servers sometimes present another argument. Server virtualization can help to insulate an organization against hardware failure.
Imagine, for example, that you have a resource-hungry application on a physical server. If you were to virtualize that server, then you probably would not be able to run any additional virtualized workloads on the host, because the virtual machine (VM) would consume most of the host server's physical resources. However, that doesn't mean it's not worth virtualizing. Virtualizing such a server could add flexibility, making it easier to move the VM from one host to another. This can help to protect against hardware failures.
Some would be quick to point out that you can accomplish the same degree of protection with failover clustering. However, clustering at the host server level is often less expensive than building separate failover clusters for each mission-critical application. A fully virtualized infrastructure also makes the cluster management process a lot easier.
Financial incentive to virtualization
By virtualizing everything, you might be able to implement a degree of consistency, while also reducing licensing costs. This concept does not hold true for every hypervisor, but it applies to Microsoft products.
Imagine that you have several servers running older operating systems that you have considered upgrading to Windows Server 2012, but have been holding off because of the cost. Virtualization can give you an inexpensive way to perform the upgrade. Windows Server 2012 Datacenter Edition allows for an unlimited number of VMs, as long as the VMs are running Windows Server 2012. A fully virtualized infrastructure could greatly reduce the number of operating system licenses that you need.
The problem with aging hardware
Some organizations are understandably reluctant to virtualize extremely old servers. Sometimes however, virtualization can breathe new life into an aging server.
A couple of years ago I had a client running a mission-critical application on Windows NT Server. Unfortunately, the application was not compatible with any newer operating systems, and Windows NT Server does not run on newer server hardware. The physical server running Windows NT was nearly 15 years old, and it was only a matter of time before it quit working.
Even though Microsoft does not officially support the idea, I was able to virtualize the server and host it on modern hardware using Hyper-V. This not only improved the server's performance, but also ended the mission-critical application's dependency on aging physical hardware.
Specialized hardware throws a wrench into virtualized infrastructure
One of the biggest reasons cited for not virtualizing servers is that some use hardware that cannot be duplicated in virtual environments. Thankfully, as hypervisor technology has improved, this is becoming less of an issue. Hyper-V 3.0, for example, supports the use of guest Fibre Channel. This makes it possible for a VM to connect directly to a storage area network.
Despite improvements to hypervisor hardware support, there are still some configurations that cannot be virtualized. For example, some applications use USB devices to enforce copy protection, but some hypervisors do not provide USB support. Even if the hypervisor did offer USB support, migrating a VM from one host to another would break the connection between the virtual server and the USB device.
Physical domain exceptions
In a Hyper-V environment, it is sometimes advisable to leave a couple of domain controllers and a DNS server running on physical hardware. This is particularly true if all of the Hyper-V host servers have been joined to a domain. Having these physical controllers offers a degree of protection in the unlikely event that something happened to the virtualized domain controllers. The physical domain controllers ensure that administrators are able to log in to Hyper-V servers to fix the problem, even if all of the virtualized domain controllers are offline.
As you can see, there are compelling arguments both for and against virtualizing all of your servers. Ultimately, each organization needs to carefully consider how server virtualization will impact its operations.