An OpenStack View on the Critical Mistake Made by Amazon Web Services and The Dilemma Facing VMware


Yesterday I posted the first part of a long interview with the founders of Piston Cloud.

In part two, Former NASA Nebula Chief Technical Architect Joshua McKenty, and Rackspace alumni Christopher MacGown trace the roots of the OpenStack movement to the 1980s and the move to convergence.

In McKenty’s view the path to convergence took 30 years. That is until Amazon Web Services (AWS) made the critical mistake of separating compute and storage. That lead to the issues with AWS early adopters who found AWS too slow, an issue caused by the separation of the two environments.

AWS influenced all the new cloud providers who followed suit by continuing the separation. This points to OpenStack;s differences with AWS and VMware.

OpenStack is combining object storage, block storage, networking and computing into one project. McKenty says the impact of convergence is similar to the impact MapReduce had on the big data movement.

It’s here where McKenty makes an interesting point. And it reminds me of a great interview I had this week with Adara CEO Eric Johnson (follow up soon).  He makes the point that the work being done in software defined networking is more important than the invention of the browser. The browser made it possible for the user to access the Internet’s resources. You could customize how you looked at your browser. In Johnson’s view, and shared by the OpenFlow community, the network becomes the execution environment. It’s the super node in some respects. The data center becomes a hyper node. The client can be optimized for whatever the demand. In essence, the network can be anywhere and customized to any need.

It means customization in a way that has the same implications for how we can control our physical surroundings. With a dimmer switch, I can turn the lights in my livingg room up or down. That same capability is coming in terms of how we customize the network to define how data behaves. In that respect, I have to agree with Johnson. The abstracted network is far more significant than the browser. If we can customize our environment through the network then we can also make data objects as much as we now have the ability to make material objects from physical resources.

Which brings us back to the question about VMware. Is the company prepared to make this leap? McKenty says no. It’s too much of a cultural shift. VMware is married to the hardware vendors, be it EMC, NetApp or HP.

And it comes down to the channel, which makes huge margins on selling the storage hardware loaded with VMware virtual machines.

Will those systems integrators want to give up the margins? Doubtful.

McKenty makes a great point but OpenStack is still far behind VMware and Cloudstack to some extent, too. AWS Marketplace, which launched this week, has the chance to tie in the channel and become the next Microsoft. They are way ahead even if the infrastructure is not converged.