More is not always good -- definitely not when it comes to hypervisors. Some IT administrators have considered building a multi-hypervisor data centre infrastructure
One of the biggest advantages of virtualisation is that hypervisors and their management tools bring data centre efficiencies once servers are virtualised. IT professionals have become proficient in designing a successful virtualisation strategy and in selecting the hypervisor that best suits their needs.
Recently, however, third-party software vendors (and even some end users) have been talking about using multiple hypervisors in the data centre. But stacking your data centre with more hypervisors may not necessarily be a good thing.
What is a multi-hypervisor data centre, and why do admins want to try it?
Cost savings is one reason more admins want to use multiple hypervisors.
Consider how having a multi-hypervisor
data centre infrastructure can be compared to having storage tiers with multiple SAN or NAS
arrays (not storage tiering built into an array). These storage tiers are, in many cases, from
different storage vendors. For example, an organisation could have an EMC Fibre Channel SAN for
tier- one apps and a Drobo iSCSI array for tier- two apps (and even a CIFS
share on a server that has JBOD
(just a bunch of disks) for its backup repository).
Organisations use storage tiers in this way to save money. The higher the performance of the storage product, the higher is the price tag. So IT chooses to pay for the best (expensive) storage tool for tier- one apps but looks to save money on the storage tool for mid-tier applications.
Likewise, multi-hypervisors in the data centre can serve the same purpose.
For instance, let's say an organisation’s IT implements VMware vSphere for its tier- one apps and Hyper-V for its mid-tier apps. The assumption is that vSphere has a higher cost than Hyper-V and IT is willing to pay the most for its most critical apps.
While vSphere platform costs more than Hyper-V, it offers advanced features such as distributed resource scheduler, or DRS, and fault tolerance, or FT). Hyper-V may cost less, but its features are just “good enough”.
Note: These are just examples of cost scenarios; your costs for these hypervisors will vary, so make sure you do your own cost comparison.
In theory, any company could have multi-hypervisors in their data centre, but usually only large enterprises or service providers implement this kind of “hypervisor-tiering”. Such enterprises have the resources and expertise to manage multi-hypervisors and could see significant cost savings by using it.
But while tiered-storage is straightforward and beneficial, implementing a multi-hypervisor strategy can be a lot trickier.
Five big problems of multi-hypervisor data centres
While it may make sense to use multiple hypervisors in some cases, here are five problems that can affect companies using a multi-hypervisor data centre infrastructure:
1. Lack of single-pane management. Although there is an increasing number of claims that
some third-party tools are beginning to support multiple hypervisors, most
of these tools still don’t and, for those that do, the feature-set isn’t the same.
If your third-party vendor does support a multi-hypervisor infrastructure, it doesn’t mean you’ll get the same visibility and management functionality. For example, in a performance tool, you can't view the same performance data in Hyper-V that you can in vSphere, and you would have trouble making any apple-to-apple comparisons).
Additionally, centralised management tools from VMware and Microsoft may be designed to support multiple hypervisors, but the capabilities just aren’t the same. Case in point: While VMware’s vCenter has a fling called vCenter XVP Manager that allows you to see Hyper-V hosts and virtual machines (VMs), you can only perform basic power on/power off on those VMs and convert them. That’s not enough if you want to use it to manage a multi-hypervisor data centre in the long run.
2. Loss of efficiency. When IT admins implemented virtualisation, they wanted to make
themselves more efficient, and they wanted greater business agility and to maximise ROI for the
company. While a multi-hypervisor data centre may maximise ROI, it actually hurts efficiency
because admins have to manage more than one platform, likely from different interfaces.
Additionally, implementing multiple hypervisors in the data centre can hurt business agility because now admins have multiple template libraries, multiple ways to deploy new VMs, and (potentially) multiple private clouds.
3. Confusion. As a norm, having more than one of something just leads to confusion. To prevent slipups, virtualisation admins need to spend more time documenting the data centre infrastructure and procedures. They can then undergo additional training on how to administer, use and manage multiple hypervisors.
4. Backup/Recovery/DR. Admins want their backup, recovery and disaster recovery plans to
be as efficient and streamlined as possible. Having multiple hypervisors complicates decisions on
which application to use to back up virtual machines, which application to use to restore them, as
well as their data centre recovery plan.
Admins have to address questions such as: Do my backup and DR tools support multiple hypervisors? If a VM fails, what application do I use to restore it? Is the potential for added downtime really worth the money I saved by having multiple hypervisors in the first place?
5. Incompatibility. There are tools that make hypervisors compatible, but the hypervisors are not compatible by default, so IT has to manually do it. Besides other tasks such as moving a VM from one platform (say, Hyper-V) to another (vSphere) and back again become very tricky. Compatibility provides greater efficiency and multi-hypervisors don't go in that direction.
Much of the talk about having a multi-hypervisor data centre started when VMware released vSphere 5 with vRAM pooled pricing. This “vRAM tax” (as dubbed by some) gave them the fervour to investigate another hypervisor and try it in their data center. While it may seem appealing at first, most companies will quickly see that administering multiple hypervisors is inefficient and not worth the potential cost savings.
If your company will see significant cost savings by switching hypervisors, then that is a
better option than having multiple hypervisors. My recommendation when it comes to hypervisors in
the data center is to use one hypervisor or the other hypervisor, not both.
David Davis is the author of the VMware vSphere video training library from TrainSignal. He is a vExpert, VCP, VCAP-DCA and CCIE #9369 with more than 18 years of enterprise IT experience. Davis is a regular contributor to SearchVirtualDataCentre.co.UK.
This was first published in July 2012