Please stop talking about private cloud

I made a presentation at yesterday’s conference on Developments with the Digital Oilfield in London. The title of my talk, “Why private cloud is a cul-de-sac of doom”, was somewhat tongue-in-cheek, and intended to be mildly provocative. However, I had a serious purpose, in that the words and terms we use to describe things are important in creating clarity and driving ideas. Misusing them dilutes their power and ultimately diminishes opportunities. In that context, the term “private cloud” is one that has minimal value and causes confusion.

In my talk, I referenced the NIST definition of cloud computing, and my version of the three key elements that embody the transformational impact of the cloud:

  • A usage-based payment model, whether that’s per user, per cycle, per cpu, or whatever
  • Rapid elasticity, or the ability to seamlessly grow and shrink your demand without needing to stop to add new hardware or software
  • No barrier to exit or entry

Given these criteria, it is hard to imagine having the need or resources to build a private cloud. Indeed, the NIST document states that private cloud applies when “the cloud infrastructure is provisioned for exclusive use by a single organization comprising multiple consumers (e.g., business units)”. This is somewhat poor, in that it masks the reality of the situation, and simply causes confusion. It also provides the opportunity for vendors to wrap up their existing technology and sell it under the new “private cloud” banner.

That confusion was already clear during the conference, when one presenter claimed that they had built their own private cloud, and another claimed to have definitely seen one in the wild. What was being described in the first case was simply a data center built using cloud technologies, like virtualisation and automated deployment, and I suspect the latter was something similar.

While I have no doubt that the use of cloud technologies can significantly improve many current data centres, there’s tremendous value in reserving the term “cloud” for services that meet the criteria I’ve outlined above. Then, when we talk about moving a customer to the cloud, we can be sure that they are benefiting from the massive improvements that such a move can deliver, rather than the incremental improvements that arise from applying cloud technologies in their own data centres.