In a few short years it seems like most enterprises have transitioned to virtual infrastructure in a pretty big way. How many of your servers are VMs now, 50%? 70%? The key enabling technology in this transition has clearly been VMware vSphere. They have been the market leaders up to this point in terms of features and stability. However, other options are becoming too compelling to ignore.
No it’s not because anyone has a “better” product. I think it’s amusing to see the comparison tables for Hyper-V vs. VMware, and how many CPUs, RAM, etc, each one supports. That’s not really the issue, nor is it a question of comparative performance. As competing products become viable alternatives, the choice comes down to overall cost and supportability.
With the recent release of Windows Server 2012 and its associated new version of Hyper-V, I consider Hyper-V a reasonable alternative to VMware, at least for Windows-based virtual workloads. I’m sure there are minor differences in performance and VM density that can be achieved on similar hardware with Hyper-V vs. VMware, but the differences probably wouldn’t amount to much. The software license costs are the bigger issue.
A proper deployment of Hyper-V certainly isn’t cheap. In order to create a highly available virtual infrastructure, you’ll need a number of OS licenses for the hosts, which would likely be data center edition licenses to provide unlimited virtualization rights and multi-node clusters, and you will need System Center Virtual Machine Manager (SCVMM).
SCVMM is part of the System Center suite, and this is where the cost equation gets interesting. Your organization may already own the System Center suite, if you use SCCM for config management or SCOM for monitoring. Microsoft makes it more attractive to purchase the entire suite rather than to buy individual components. If you own the suite, then you’ve already got the rights to use SCVMM. If this is your situation, then it may well be cheaper to deploy Hyper-V than vSphere.
All that said, the list of supported Linux distributions that run on Hyper-V is very short, including only the major RPM-based distros: Red hat, CentOS and SUSE. If you want to run Ubuntu, BSD and others, you may be out of luck, or at least unsupported. This doesn’t mean that Hyper-V should be off the table, it just means that we need to consider having more than one hypervisor.
If you run Red hat Linux, and you want a commercially supported hypervisor, Red hat Enterprise Virtualization (RHEV) may make sense, for the same reason that Hyper-V makes sense for Windows. Licensing the OS on the underlying host processors can make the solution cost effective.
RHEV is based on the KVM hypervisor, and KVM can be deployed for free on free Linux distros like CentOS and Ubuntu (and many others). KVM is the defacto standard hypervisor for OpenStack, which looks soon to become a very popular private cloud solution. Another open source alternative hypervisor is Xen, in use by Amazon in their public cloud.
I suspect that over the next few years, we’ll see hypervisor market share shift away from VMware and more toward Hyper-V and KVM, and we’ll likely see specific workloads aligned to each hypervisor based on what OS is running on the VMs and what level of support is desired by the enterprise. For example, I’d think that business-critical linux-based applications would be deployed on VMware, Windows workloads deployed on Hyper-V, and disposable cloud instances deployed on KVM, via OpenStack.
This will be a significant shift for some enterprises that traditionally maintained Windows and VMware skill sets. There will be an increasing need to know Linux, especially at the infrastructure level. By this I mean the things we think of when building a virtual infrastructure. We’ll need to know how to configure multiple NICS, VLANs, multi-path SAN storage and the like, under Linux. My recommendation? If you haven’t already, get busy learning Linux!