The multi-hypervisor fallacy

IT is at a data center crossroads. Virtualization is going to happen; organizations can choose to either deploy it strategically or tactically. Strategically virtualizing the data center transforms it into a private cloud with automated self-provisioning including metering, monitoring and chargeback. Tactically virtualizing servers incurs unnecessary inefficiencies and lost opportunities. Embracing multiple hypervisors is emblematic of the tactical approach.

The Downside of Multiple Hypervisors

Microsoft touts the ability to easily manage "all Hyper-V and VMware virtual machines" as the key differentiator of its virtualization platform. The messaging may be having an impact. A February 11, 2009 article by Alex Barrett claims, "For better or worse, an increasing number of IT managers run a mixed virtual server environment that includes VMware and Hyper-V." A September 2009 Hot Aisle reader poll showed 44% of the respondents use two or more hypervisors.

Microsoft MVP and VMware vExpert Greg Shields is a self-proclaimed multi-hypervisor advocate. I've collaborated with Greg over the years on several articles and white papers, but we disagree about mixing vSphere and Hyper-V. In a recent podcast on the subject, Greg provides a rationale for the increasing trend, "Some workloads need very high-end virtualization. Other workloads can use general-purpose virtualization and we can choose what makes sense for us from a cost perspective."

While on the surface this might seem sensible logic for an economically minded CIO, the overwhelming savings and other benefits resulting from data center virtualization make the delta in hypervisor costs insignificant in comparison. And hypervisors are not deployed in isolation; different equipment and administrative requirements can quickly negate any perceived price advantage.

Implicit in multi-hypervisor advocacy is an undertone of virtualizing servers rather than the data center. This myopic perspective limits both savings and synergies. Cisco studies, for example, show a lack of vNetwork capability results in 30% fewer servers that can be virtualized along with 30% higher administrative requirements. Network administrators have no way to monitor traffic over a vSwitch for compliance, auditing and troubleshooting purposes, and they cannot apply network and security policies that follow a VM as it live-migrates. Since only vSphere enables vNetwork capabilities, multiple hypervisors leave at least a portion of the data center running less efficiently and less secure.

Shield's January 7, 2010 commentary describes the all-important management role in a multi-hypervisor environment:

"Consider today's hypervisor playing field: On one hand, you've got high-end capabilities with high-end price tags. On the other hand, you've got general-purpose virtualization that in many cases has no price tag at all. Now, combine this with the new recognition that not every workload has the same virtual infrastructure requirements, and you can see how multi-hypervisor management tools become a brilliant play. Let's implement the right one in the right circumstances, and layer a pervasive multi-hypervisor solution over the top to manage them all from a single pane of glass."

The flaw in this argument is that even Microsoft's System Center Management Suite does a poor job of managing VMware vSphere; it is quite unlikely that smaller manufacturers will be able to create effective products. VMware focuses on optimizing resource pool management while enhancing cloud attributes ranging from agility to compliance with an arsenal of virtual infrastructure administration and automation tools.

Multi-hypervisor advocates like to point to the mixed server operating systems of a traditional data center as evidence that a heterogeneous environment can thrive. But the propensity of disjointed technologies, equipment and processes is certainly not planned – it's driven by disparate application and departmental projects. Virtualization, when deployed as the underlying platform, enables data center unification. Utilizing multiple hypervisors, on the other hand, not only perpetuates the inefficient status quo, it adds yet more islands of technology to manage and coordinate.

Transitioning a Virtualized Data Center to a Private Cloud

Microsoft says that it "…takes a broad view of management, with virtualized technologies given equal weight to their physical counterparts", and that physical machines will remain a "key part" of IT infrastructures. Organizations adhering to Microsoft's messaging tend to discover, upon reaching a certain virtualization threshold, challenges containing virtual machine sprawl and efficiently provisioning compute, network and storage resources. They continue to face the myriad inefficiencies and high costs of operating a physical environment, compounded by requirements to also manage hypervisors, VMs, hosts, vSwitches, etc.

A completely virtualized data center is a key to the private cloud where standards, policy-driven management and SLAs enable streamlined IT as a service. Even organizations leisurely strolling down the virtualization path still benefit as long as they keep the big picture top of mind. They reduce or eliminate unnecessary expenditures on physical data center vestiges of servers, PDUs, air conditioners, etc. More importantly, they make architectural platform evaluations of network, compute, storage, virtualization software, disaster recovery and desktop within the context of their contribution to the private cloud end-goal.

Virtualization may not be a magic bullet technology, yet it unquestionably is revolutionizing the data center. The overwhelming response to Cisco's UCS shows that organizations understand the value of a compute platform optimized to host virtual infrastructure. The unprecedented collaboration of multiple leading manufactures on VCE vBlocks and Secure Multi-Tenancy also indicates a major shift in data center approach. The ability of IT organizations to pull together both the internal consensus and monies to purchase these cross functional solutions validates the "virtualization changes everything" mantra.


Author Disclaimer: I work for a leading VMware partner, but the opinions expressed in this article are my own and are not approved or endorsed by my employer.






This entry was posted in Uncategorized and tagged , , , , , , , , , , , , , , . Bookmark the permalink.

3 Responses to The multi-hypervisor fallacy

  1. nate says:

    Low cost? If you want low cost vmware has options, maybe people forget
    Sad that so many immediately gravitate towards the highest premium version of vSphere when they think of vmware and say “oh, it’s too expensive!” Well duh, don’t use that version.

  2. Steve Kaplan says:

    That’s a very good point. I didn’t want to include it in the article since it’s all about taking a strategic approach to virtualizing the data center, but the reality is that there will be some situations in many organizations where not all applications will be required to run as VMs on the primary data center virtualization platform. As you mention, though, it makes a lot more sense to at least use the same family of hypervisor platform.

  3. Ruudt says:

    but the reality is that there will be some situations in many organizations where not all applications will be required to run as VMs on the primary data center virtualization platform.

Leave a Reply

Your email address will not be published. Required fields are marked *