Konstantin Emelyanov - Fotolia
Multi-hypervisor environments can introduce complexity and management challenges, but they can also provide flexibility and cost savings.
When virtualization first gained traction in the early 2000s, IT administrators only had a few choices: VMware and then other products, such as Virtual PC, that weren't as reliable and didn't offer as many features. This led to mammoth growth in large VMware shops that pushed VMware's dominance in the data center to unprecedented levels.
As time passed and the hypervisor grew into more of a commodity, as opposed to a unique data center feature, other hypervisors emerged: Microsoft's Hyper-V and the open source KVM. For the dedicated VMware admin, even considering another hypervisor could be considered traitorous, but if you take a look at everything with an open mindset, you might find there are key use cases for multi-hypervisor environments.
Three major considerations you should take into account when evaluating virtualization platforms, even before the technical aspect, are workload types, use cases and cost. Use cases and cost specifically go hand in hand when deciding on hypervisor platforms.
Consider production, test/dev and infrastructure workloads
Let's start with the workload types you most likely have. Many companies run production, test/dev and infrastructure workloads. Although some workload types vary, it's likely many of your servers fall into these three categories.
Workload type is only part of the equation, though; you must consider workload importance as well. People typically assume production servers are the most important, but critical infrastructure servers, such as Active Directory controllers, can have the same or a higher value than a production server. Understanding what your workloads consist of is critical when reviewing different hypervisors.
Compare performance and reliability
This is where things get a little controversial. For a baseline, we must establish an order for the major hypervisors. VMware has a long track record of being the dominant hypervisor and is proven in the data center. Hyper-V and KVM are both maturing, but Hyper-V has more resources behind it. When it comes to performance, reliability and cost, VMware is at the top, with Microsoft in second and then finally KVM. This doesn't necessarily mean one hypervisor is better or worse than another; it's simply a starting point for evaluation.
Now that we have a baseline, we can look at the workload types and where you might place them. You don't need three hypervisors in-house; two should be more than enough. The key is to consider at least two different platforms based on the criteria you set.
No one wants systems to go offline, but the simple fact is that they can and will. What varies is how often systems go down and how long they're down for. For systems that can be offline for longer periods of time -- such as test/dev, back domain and network servers, logging servers and monitoring servers -- it makes sense to use a hypervisor platform that's a solid performer but might not have every bell and whistle that a top-tier hypervisor has.
Weigh the potential risk against cost savings
The main reason to adopt a multi-hypervisor environment is to save money. A top-tier virtualization platform, such as VMware, also comes with a top-tier price. It's not that the price isn't justified. It's more about whether you need everything that VMware can do for you at all workload levels. Often, the answer is no.
Test/dev servers are typically light duty and have limited runtime due to only being used in upgrade cycles or sandbox environments. These aren't the workloads you want in your top-tier virtualization platform. A secondary virtualization platform enables you to save more of your higher-end resources for more critical production applications. This approach gives you the best value for your dollar. It's an introduced risk, because lower-tier virtualization platforms might not be as extensive as top-tier platforms, but it can be an acceptable level of risk when balanced against the cost savings.
When you know what workloads need what hypervisor and why, it's easier to figure out virtualization licensing. Typically, virtualization licenses come in two flavors: from the core production higher-tier value to the lower-cost or even free hypervisor. Saving on hypervisor costs can be a game changer for your budget, so the ability to tier and embrace a multi-hypervisor environment can be a huge win. This is both in terms of the raw budget savings and the best value for your software license purchases.