The May issue of "Communications of the ACM" contains an article that examines the similarities (and the significant differences) between cloud computing and electricity generation. The analogy between the two was famously made by Nick Carr in his book "The Big Switch." In his book, he posited that IT operations would move into large-scale, centralized "clouds" because of economies of scale, much as electrical generation became the province of utilities in the 19th century, making on-premise power generation obsolete.

The ACM article agrees -- to a point -- with Carr's thesis, but notes several caveats.

First and foremost, the authors point out an obvious difference between power and computing -- power technology evolves quite slowly, while the pace of change in computing, driven by Moore's Law, is staggeringly rapid. I made this observation in my first book, in which I criticized Carr's comparison of railroads to IT in his "Is IT Important" book. The fact is, the pace of improvement of computing is -- literally -- unprecedented in human history. It outstrips our ability as a species to comprehend. One can expect cloud computing to evolve at a similarly dizzying pace. This is why we (HyperStratus) do not agree with the various efforts to "standardize" cloud computing -- it's far too early in the cloud's evolution and innovation is proceeding apace. Standardization makes sense when the innovation of a particular offering has been wrung out. We are nowhere near that stage with cloud computing.

The authors then turn to what they term "The Limits of Scale" and make the observation that not all computing tasks are suitable for parallel processing. They then go on to state that, for many companies, the problem of scale can be avoided by implementing private clouds, noting that current developments in power generation seem to moving to decentralization via wind farms and micro-turbines. Without getting into the whole private cloud debate, I would note that there are many non-parallel compute tasks that can still take advantage of the horizontal scalability of cloud computing, and that network and storage provide other reasons to use a cloud provider. I think the "move to decentralized electrical generation" analogy is a red herring, in that very few of the alternatives being explored will prove to be economically or technically feasible.

The authors then turn to what they characterize as "business model" issues with cloud computing, citing complementarities and co-invention, lock-in and interoperability, and security.

The first issue is really nothing to do with cloud computing; the authors note that companies that merely use cloud computing without exploring how it can change the type and design of their applications will not garner all the benefits of cloud computing. This is really an issue for users of cloud computing; a different way of saying this might be "cloud computing enables companies to do computing tasks and create applications that previous computing modes prevented. It's a tremendous platform for innovation, which users should seek to leverage." I recently addressed this in my CIO blog, in which I stated "Cloud computing does not equal Hosting 2.0."

As far the topic of lock-in and interoperability, my position is that lock-in inevitably accompanies early use of a new technology. Gaining the benefit of a new generation or new technology platform requires use of what Clayton Christensen calls "an integrated offering," meaning all of the various components come from a single supplier which integrates them. Waiting until lock-in and interoperability are "solved" will condemn an organization to falling behind in the race to the cloud. Lock-in is the flip side of innovation, so it makes sense to acknowledge it and architect to reduce its power.

Turning to security -- the last issue the authors list as a "business model" problem for cloud computing, I believe this is a short-term problem likely to be transitory. Security is the issue regarding cloud computing -- today. I don't regard security as an inherent failing or shortcoming of cloud computing, and believe that the current concern with security will moderate over time. It's human nature to believe that new practices are very risky, and to underestimate the risks associated with current practices. Long-term, it's likely that cloud computing will, in general, be favorably evaluated on the security dimension when compared to on-premise practices.

The ACM article is a great contribution to the discussion of "cloud is like electrical generation utilities" and brings a more nuanced evaluation of the simile. The authors provide additional perspective on the comparison without negating its usefulness. I definitely recommend reading it.

• Bernard Golden's blog
• Add new comment