What a mix-up: using different hypervisors

When asked to write about the pros and cons of deploying multiple hypervisors, my first thought was that nobody in their right mind would want to do such a thing, especially if the software was from different vendors. Management would be a nightmare and so would all kinds of compatibility issues.

On further investigation, it starts to make sense, and even turns out to be fairly common.

Forced conversion

Here are a few good reasons why you might end up having to manage more than one hypervisor.

1. You have merged with another organisation running a different hypervisor, perhaps from a different vendor. Forcing a change would mean re-training staff and converting virtual machines, with consequent downtime and possible loss of functionality.

2. The applications you want to run fit a particular hypervisor technology. For example, companies with Citrix application servers may find they work best when hosted by Citrix XenServer, while Microsoft applications might deliver more when run on Microsoft’s own Hyper-V.

3. Features that are important to your applications may be available only with a specific hypervisor – massive scalability, for example, or the ability to dynamically assign and balance resources. Live migration of virtual machines may also be critical, or the ability to take snapshots and quickly recover crashed systems.

4. It can make sense financially to mix hypervisors. Most companies can find the budget to go for the best there is for mission-critical applications, for example, but for others a freebie may be all that’s allowed.

Welcoming diversity

Whatever the reason, hedging your virtualisation bets with multiple hypervisors does not seem such a big challenge – no harder, in practice, than mixing together physical servers from different vendors. Plus there can be real benefits.

1. Given the right hardware it is possible to host hundreds of virtual machine on a single server, but spreading workloads across multiple hypervisors on multiple servers can improve both performance and availability.

2. Hypervisors vary widely in features and functionality. Limiting yourself to a single implementation could limit opportunities to deliver best-fit solutions to users and customers.

3. Multiple hypervisors allow for clear separation of both virtual and physical environments, eliminating the risk, for example, of a development virtual machine crashing another running a production system. Microsoft and other vendors do support competing systems.

4. As well as separate operating environments, multiple hypervisors allow for clear demarcation of management responsibility.

5. The hypervisor market is still evolving. Vendors and products change all the time, so it is prudent not to lock into a single vendor before the dust has settled.

Of course there are issues to bear in mind, not least the need for some kind of integrated management solution able to handle all your hypervisors from one central console.

Not getting along

A number of products are appearing that can do this but they are in their infancy. The platforms they support and the level of functionality afforded to each one is limited – although Microsoft is certainly making an effort to address this with support for Hyper V, Xen and Esx, among others. Rival vendors are following suit.

Interoperability is another concern, as there is no single standard for virtual machine and disk formats. Conversion takes time and can throw up incompatibilities that are hard to resolve.

Lastly, it is important not to forget one of the key benefits of virtualisation: efficient use of hardware resources. Multiple hypervisors spread across separate hosts can dilute that benefit and make it harder to apportion physical resources where they are needed most. ®