VMworld 2012 San Francisco is over, and I have some time to reflect on my virtualization thoughts in general before getting ready for VMworld Barcelona. One thing I noticed is recent announcements about VMware vSphere 5.1 and Microsoft Hyper-V 2012. Microsoft and VMware both released a specific new feature to each respective platform at basically the same time. Is this a sign that Microsoft is really closing the gap on VMware? I think we are getting there, but I have also made some other personal observations on how I see virtualization in the future, and I foresee a completely different method and mindset for the future between these two companies.
The recent “feature” or “enhancement” called, for lack of a better term for both platforms, a share-nothing migration, was released in both vSphere and Hyper-V. This feature takes away the need for shared storage to perform live migrations from host to host. This is a huge win for the SMB space as well as an added bonus for anyone trying to piece together any kind of lab system. Is this really new for VMware? No, they have had storage migration in place for a little while now, but have now removed the requirement of shared storage for vMotion.
When I was reading through information for both vSphere and Hyper-V I had an interesting observation on how differently Microsoft and VMware seem to visualize the future of virtualization. They both have very different approaches to the ecosystem as a whole. Microsoft has recently released Microsoft System Center 2012 and is marketing this “as the only unified management platform where you can manage multiple hypervisors, physical resources, and applications in a single offering, versus multiple fragmented point solutions delivered by the competitors.” Let’s look at this: Microsoft is dropping the idea of a suite of related but separate products and creating a single unified management tool for servers and hosts alike.
VMware has been moving in an entirely different direction with its suite of related but separate appliance-based virtualization management offerings, a collection of different VMware appliances that perform specific tasks, but where the data can be pulled into vCenter to be able to manage the environment from the single pane of glass called VMware vCenter. vCenter itself has already been ported to an appliance, and it seems to be just a matter of time before the appliance will be the only option available. This will likely be true as well for any new features or services released from VMware moving forward.
I can see the pros and cons in both the unified method and the appliance suite approach and will be curious as to what your opinion might be on what you think is the best path moving forward. I still believe that VMware’s top priority is in the larger environments, where VMware has a greater footprint; I can picture an ecosystem running on several well-oiled appliances working separately but also in harmony with each other. Once VMware is completely an appliance-based eco-system, thus removing the prerequisite of using Microsoft Windows for any of its tools or solutions, it will become, in my opinion, a completely separate part of the infrastructure like application, network, midrange, etc.
Microsoft, on the other hand, does not want to define virtualization as anything but an extension of the current Microsoft ecosystem, where all systems live in harmony within System Center. That concept will really appeal to the SMB space, where there are more and more administrators that are required to “maintain anything and everything” in their environments. This is one of the reasons that I think Microsoft is going to do really well with virtualization in the SMB space, and I think VMware will continue to command the larger enterprise space.
In closing, I have one other thought to consider for the future. In the past decade, virtualization has gone from the lab to running close to 50% of all compute workloads. Using the past decade as a reference, one could conclude that quite possibly, in a decade from now, virtualization could be responsible for close to 100% of all compute workloads. If that works out to be the case, then wouldn’t it indicate that with the potential for virtualization moving forward, virtualization should be a separate, isolated part of the infrastructure?