The mere mention of the phrase ‘multiple server virtualization hypervisors’ can send many administrators into a panic frenzy. The daunting complexity of such an undertaking combined with the sheer overhead required to maintain multiple hypervisors would rightfully stress even the most talented administrator. But the reality is that if a company is looking to refresh hardware from a server virtualization implementation or looking to virtualize more of their infrastructure, it’s an important topic to discuss – even if the conversation requires using a brown paper bag.

The truth of the matter is that managing a multiple hypervisor environment is challenging, but not impossible. When administrators stop and take the time to think it through, it is no more complex than maintaining multiple physical hardware form factors and vendors in a datacenter or configuring and managing disparate technologies to deliver desired workloads. 

The key to lessening stress levels is upfront preparation, in short, understanding the use case and business justification for migrating to a multiple hypervisor environment. Typical reasons for utilizing multiple hypervisors can span from specific environment requirements, application support, and performance optimization to a need for single vendor support, or tapping security or high availability features within a hypervisor. Regardless of the reason, administrators should perform an internal audit to review applications, existing workloads, storage requirements and hypervisor feature requirements, to help their company make an informed IT business decision.

The following questions are a sampling of things administrators should evaluate before making a decision.

1. Do I have applications that would gain better support or performance by running on a different hypervisor?
2. Do I need features or functions that are not available with my existing hypervisor?
3. Are there new system requirements or virtual machine requests that would be served more effectively or cheaper on a different hypervisor?
4. Are there disaster recovery or business continuity requirements for particular virtual servers that can only be provided by a different hypervisor?
5. Are there security or reporting requirements that are not met by my existing hypervisor?

If any of these questions piqued your interest, then perhaps a multiple hypervisor environment might be next on your checklist.

Making the case

There are various scenarios where managing multiple hypervisor vendors in a server virtualization environment makes sense. However, the most compelling reasons center on application vendor support or performance and separation of environments. 

A poll conducted by Enterprise Strategy Group analyst Steve O’Donnell¹ indicates that application vendor support, mitigating the risk of not getting support because of the hypervisor your virtual server is running on, is one of the leading reasons that multiple hypervisors are utilized. Often an application vendor will only support their application on a virtual server if it is running on a vendor-approved hypervisor. Another challenge is application performance and/or optimization. If a company opts to run applications on a non-vendor approved hypervisor, companies run the risk of forfeiting performance increases, built-in optimizations and even application features - a known frustration for many administrators.  

A multiple hypervisor environment eliminates these challenges and provides opportunities to optimize performance and streamline operations. For example administrators can use a Citrix XenServer to run Citrix XenApp or Microsoft Terminal Servers.  A 10 server farm of XenApp servers on average can achieve a 2 – 3 server reduction by running on a Citrix XenServer instead of on a designated vendor platform. Similarly a virtual desktop infrastructure (VDI) environment utilizing memory over-commit features from VMware may realize significantly higher VDI density rates than the same workload running on a designated vendor hypervisor without memory over-commit features. The question then becomes ‘Am I getting the most out of my existing infrastructure and vendor application support?”

Another compelling reason to consider multiple hypervisors is separation of environments. Administrators can more efficiently separate or partition virtual environments to meet departmental budget requirements or distribute administrative resources. For example, a desktop management group can run their VDI environment under a separate hypervisor than the server management group without encountering performance or access issues.  A server administration group can give developers a hypervisor that they can more easily manage and maintain themselves thereby reducing the server management team’s workload. While these resource allocations can be achieved in a single hypervisor environment, the complexity of managing those additional features can create as much overhead as running a second hypervisor, but without the added benefits of having another hypervisor at one’s disposal.

While these are only two examples, this type of setup can provide a litany of customer specific and/or application benefits that should be considered before dismissing the idea because it is perceived as “too complex.” 

Below are several reasons to consider a multiple hypervisor environment, including:

• Reduced overall cost of server hypervisor software
• Improved application compatibility and support
• Some server workloads will perform better on particular hypervisors
• Administrative requirements for non-server administrators
• Single workload deployment on appropriate hardware for disaster recovery
• Disaster Recovery and Business Continuity for specific workloads to different cloud vendors
• Vendor diversity for agility and elimination of vendor lock-in
• Clear separation of environments for different departments and/or solutions
• Single vendor support for hypervisor, operating system and applications

Reality check

However, just like anything in life, there are two sides to a coin. Same philosophy applies in this scenario. Running multiple hypervisors has drawbacks and challenges that need to be considered before making a decision. 

For example, the primary challenges include:

• Centralized management: While several solutions are currently available or coming to market to address this issue, they all are limited in feature and compatibility. As a result, the lack of a mature centralized management solution poses the greatest challenge to long term acceptance of multiple hypervisors.
• Server and storage inefficiency:  Poorly designed or provisioned multiple hypervisor environments can result in idle or underutilized servers or storage forcing companies to have to purchase additional servers or storage to properly support their needs. 
• Management and maintenance:  Management and maintenance of separate systems, including updating and upgrading components within systems are often challenges within a single hypervisor environment. That task is multiplied in a multiple hypervisor environment.
• Lack of interoperability:  The lack of interoperability between hypervisors makes each hypervisor environment a separate island of resources which cannot be grouped together.  Some vendors are working together to improve interoperability to make transition between hypervisors as seamless as possible.

While, there are certainly other challenges that are not discussed here, the advantages and flexibility of multiple hypervisors often create a compelling case for diversification.

To switch or not to switch, that is the question

While each IT department has to weigh the uniqueness of their situation and business IT goals, the advantages gained through utilizing multiple hypervisors are compelling enough to persuade even the most overwhelmed administrator. Advantages such as reduction in software cost, improved application support, lower overall virtual machine count, and logical separation of environments overshadow any increased complexity or management overhead.  There are certainly drawbacks to utilizing multiple hypervisors but the drawbacks will decrease over time and a well designed architecture can minimize many of the challenges as well as reduce the total cost of ownership for a server virtualization environment. 

With that said, let me leave you with this last thought. A hypervisor poll conducted by Steve O’Donnell¹ on The Hot Aisle blog in September 2009 indicated that 44 percent of those that responded to his poll were running two or more hypervisors. The reality is that the number of environments running more than a single hypervisor will continue to increase as more solutions continue to mature and competitive pressures continue to change the landscape of server virtualization. The question is will your company be prepared for this inevitable shift?

Ben Reeser is the lead architect for the Virtualization and Storage practice at 3t Systems, a leading IT consulting firm based in Denver, Colo.  He brings more than 14 years of experience in helping companies analyze their technical and business needs to select the right solution to meet their needs.

References

¹O’Donnell, S. (2009, September 28). Hypervisor Poll. Message posted to http://www.thehotaisle.com/2009/09/28/hypervisor-poll/