The Software Defined Data Center: That was pretty much the biggest takeaway from this year’s VMworld in San Francisco. VMware made announcements about the new vSAN that will be coming out soon to enhance software defined storage and about the NSX platform that addresses one of the final hurdles on the path to finally having a completely software defined datacenter, network virtualization. There have been plenty of write-ups on these topics, including one very good post from one of my colleges, Bernd Harzog. I am not going to go into any details on those announcements except to say that VMware is expanding and putting themselves in a good position to be the center of the virtual universe. I believe that it will take some time for software defined networking to really take off. My gut is telling me that it will be a slower process at first, just like the adoption of server virtualization, but when it does really take off, I believe the end result will have the potential to have just as great or even a greater legacy than server virtualization has.
Has VMware kept their eye on the ball during this march to the software defined datacenter? In my humble opinion, I think there is a specific ball that VMware has completely avoided: true multi-hypervisor management. Let’s face it; the hypervisor in itself is just a commodity. The real prize is the management of the complete infrastructure. Yes, VMware has released a multi-hypervisor plug-in, but as of now, it only works with Microsoft Hyper-V 2008 and provides just the basics. VMware seems to be more focused on working with public cloud companies to integrate with vCloud as well as providing the vSphere Hybrid Cloud to present as an extension of your own datacenter. I believe we have a little ways to go before we can have seamless hybrid transformation between the public and private cloud services.
If you take a look at the technology from Hotlink, this is exactly what I would want and/or expect with multi-hypervisor management. I am leaning towards the idea that this kind of integration would be better left to third-party vendors instead of OEM providers. I really think we, the consumers, will end up with a much better product with many more capabilities in the long run.
I believe we still have a ways to go before we have true heterogeneous environments, but I do believe they are still coming. Back in the 1980s and 1990s, Microsoft thought it would dominate and totally control the server market in the datacenter, but no matter how hard they tried there were these pesky things like Unix, Linux, and such that seemed to get in the way; not only did the adoption of these non-Microsoft products not go away, but they continued to grow and expand in the datacenter.
They say that history tends to repeat itself and I believe that this is going to be another example of that. As much as Microsoft and VMware want to be able to lock each other out of the datacenter, that is just not going to happen. With that said, I believe the true keys to the kingdom will be the heterogeneous management of all hypervisors with the addition of other technologies like OpenStack and CloudStack. Are we striving for a single pane of glass for all things virtual, both great and small? This is the future that I dream of in my software defined datacenter.