CloudU Notebooks is a weekly blog series that explores topics from the CloudU certificate program in bite sized chunks, written by me, Ben Kepes, curator of CloudU. How-tos, interviews with industry giants and the occasional opinion piece are what you can expect to find. If that’s your cup of tea, you can subscribe here.
I recently read a well reasoned post by HP’s Cloud CIO Christian Verstraete. In the post Verstraete reflected on a Twitter exchange he’d had responding to the claims by someone that private cloud was a myth. The private/public debate has raged for a few years now and the two sides go something like this:
To put it into context, my definition of cloud uses the acronym OSSM. That is, to be considered cloud an asset should be on-demand, scalable, self-service and measured. As long as it meets these requirements, I’m relaxed about calling it cloud. So a virtualization product on its own, for example, isn’t cloud because it doesn’t deliver on the self-service and measured parts of the requirement.
My key requirement for clouds is that they deliver on self-service and metering. My belief is that over time organizations will move to a model that seeks to very accurately identify IT costs, and report them alongside revenue figures. In order to do so, it is important that any service utilized is measurable on a per-unit basis. So, no, virtualization does not equate to cloud. But a coherent strategy of virtualization alongside automation, self-service and scale is cloud, even if it happens to occur on-premise.
This part of the discussion tends to drop into a semantic argument. Opponents of private cloud suggest that a fundamental requirement of cloud is the ability to scale and that, in the event that infrastructure resides on-premise, that scaling is therefore limited. I believe that this view is shortsighted. All infrastructure, after all, has limitations. It is conceivably possible to max out the total infrastructure available from any cloud vendor. Sure, it would take an unbelievably massive workload to do it, but it’s possible – cloud, after all, sits on real hardware somewhere. The key thing here is that for an enterprise scale means the realistic scale requirements of an individual business unit. If I run a particular department within BigCo, I don’t need to know that my company’s infrastructure can scale to map the human genome (unless of course that’s the particular requirement I have). All I need to know is that it can meet my particular requirements. Across an entire organization the kind of benefits that public cloud vendors can enjoy from different customers having usage and peak requirements are also applicable – it’s unlikely that every department within BigCo has its peak infrastructure requirement on the very same day. And if they do, that is an IT task to manage – for my business unit, I just need to know that my little marketing workload can scale as I need it to.
Yes, a fundamental part of cloud is that a user can, by simply swiping their credit card, spin up servers at will. There is no fundamental reason why business units can’t enjoy the same level of empowerment within an enterprise and using on-premise hardware. Yes, it requires some technology to do so (some automation, some charge-back mechanism), but it is possible. While I agree wholeheartedly that virtualization doth not a cloud make, but it’s going a step too far to say that it is impossible to introduce aspects of on-demand and self-service provision within the on-premise data center.
Cloud is a delivery mechanism for technology. As with any delivery mechanism, there are several ways of doing things and not every solution will fit into one rigid definition. Private cloud is a valid concept, and is delivering real value to organizations today as they strive to attain higher levels of agility within the constraints of their regulatory, compliance and internal culture landscape. To deny private cloud is to create an elitist and ultimately damaging schism in something that should be promoted.