Having read (hat-tip Dennis Howlett) Randy Bias' article at Kendallsquare on Debunking the "No Such Thing as a Private Cloud" Myth I have to say — rather like the apocryphal Irish direction-giver — if I'd wanted to make a case for private cloud, I wouldn't have started from there. Randy and I joined a civilized conversation a few weeks back as a follow-up to my earlier post on this topic, and I fear he's already forgotten every dam' thing I said. So I guess I'll have to reiterate it.
But first, let me (shock, horror!) make the case, such as it is, for private cloud. It looks like we're stuck with the term, along with all the ugly implementations that are going to be classed under it and which I fear will ultimately lead to its becoming discredited — unfairly dragging the reputation of true cloud computing through the same mud as it does so. As you can see, I still distrust the term, mainly because it is so open to misinterpretation, and I shan't be using it myself without heavy qualification and many caveats — which, by the way, should make for an interesting panel discussion with Verizon, IBM and others that I'll be moderating at the All About Cloud event in San Francisco this May [see disclosure]. But I do see circumstances where it's possible to make a case for implementing cloud-like infrastructure in a private environment, and Randy, despite starting his exposition from completely the wrong starting point, does end up making a statement about private cloud with which I can heartily agree:
"The private cloud model is a critical transitional step. It is an essential component to help larger organizations move their compute capacity to the public cloud."
The thing that cloud purists and evangelists are too prone to forget is that most enterprises are heavily committed to existing investments in pre-cloud, on-premise infrastructure. These are assets they simply can't afford to throw away or retire just yet. Very few organisations are lucky enough to be able to start over with a clean sheet and move everything to the cloud in one fell swoop. Therefore, for the next few years, the vast majority of them are going to have a hybrid IT infrastructure — some of it in the cloud, some of it not. This was an important takeaway, by the way, from my interview with SAP CTO Vishal Sikka, which I published recently. They're going to need a way of bridging the two, and that's where some kind of cloud-like private infrastructure may come in useful, to mediate between what's already in the cloud and the other IT assets that are either transitioning towards the cloud or remaining on-premise.
Where Randy and I fundamentally disagree, however, is in our interpretation of the words 'private cloud' and in what we each regard as the key characteristics of this transitional infrastructure (nor am I prepared to join him in dignifying the notion with the status of a 'model'). Randy insists that cloud computing is essentially a business model (pay-as-you-go outsourcing) built on top of an architectural model (shared virtualized infrastructure), and completely ignores — no worse, attempts to deny — that it has anything to do with the Internet.
Yet in my view, the most important attribute of the cloud — too readily overlooked by many commentators — is that it lives in the Internet. The Internet dimension is crucial because it brings with it an obligation and a necessity to remain open to connections. It means that a cloud has to have:
- Open APIs
- Unlimited bandwidth
- Collective scrutiny and innovation
The third of these is probably the most difficult to grasp and yet the most far-reaching in its impact. Any infrastructure or application service that lives on the Web as a shared resource is constantly tested by two separate yet complementary schools of users:
- Skeptics that don't trust it
- Enthusiasts that want to push the envelope of what's possible
Those two interest groups have a virtuous push-me, pull-you effect on the provider's infrastructure or application that ensures that it's constantly staying up-to-date both with every threat that might bring it down and with every emerging enhancement that could make it better. These dual competitive forces impel a cloud platform to evolve in ways that private platforms can never cost-justify. Anyone that designs a perfectly state-of-the-art cloud platform and deploys it to a private environment — even if that private environment is shared by thousands of distinct user organisations (and that's a tiny minority case) — cuts it off from the competitive pressures that ensure it continues to evolve and protects it from gradual yet inexorable decline into obsolescence.
Therefore, the only use case that I believe makes sense for private cloud is one where it acts as a temporary transition chamber. Either as a controlled environment where IT assets can be prepared for subsequent deployment to a fully cloud existence, or to mediate between public cloud assets and those left operating within the private enterprise environment. A private cloud that helps IT assets move towards the public, Internet-immersed cloud, I can live with. Anything that's designed instead to somehow avoid connecting to the wider Web is just missing the point.