Sunday, 6 February 2011

Cloud Computing. Are Utility and Cloud Computing really analogous to traditional utilities?

The analogy is often made between utility computing and other ‘traditional’ utilities such as electricity, gas, water, and telephone. The same analogy is now being made with cloud computing. At a high level this makes sense. In the same way there is no need for each individual or company to operate their own power plant, do they really need to operate their own computers? Why invest in the capital equipment to provide a computing capability if it can be piped into the premises and purchased ‘on demand’? The opportunities for economies of scale, the centralization of expertise, increased reliability and scalability all make perfect sense.

However, as Nicholas Carr also recognizes in his book “The Big Switch”, analogies often break down once you move to another level of detail as highlighted in the table below. Here we need to understand the difference between the utility (e.g. electricity) and the application of that utility (e.g. cooking).

Factor ‘Traditional’ Utility Utility/Cloud ComputingImpact
Co-location of capabilities Whilst the utility service might be produced ‘off-premise’, its application is very much ‘on-premise’. That is, the electricity might be produced at a power plant, but it powers my oven in my home. Nor do I need visit the nearest water treatment plant to take a bath. (Putting the notions of restaurants and public bath houses to one side...) Both the utility and its application are ‘off-premise’. That is, not only is the raw utility – compute and storage – located ‘off-premise’, but so are the software applications and data hosted on it. Disruption in supply means also a disruption of application. Not only is supply of the utility dependent on a 3rd party, so is its application. I can’t swap to an alternative utility source (e.g. batteries), because I have lost the application as well. How this risk compares to having one or both on-premise requires assessment.
Portability Though utility suppliers might be monopolistic in some countries or states, the application of it (e.g. powering my oven) is pretty much portable between them. Utilities all supply the same raw product. Whilst voltage might vary by continent, electricity is electricity. If I change my electricity supplier at home I do not have to buy a new oven. Other than in its most basic form – an empty un-purposed server – many utility/cloud capabilities vary from supplier to supplier, even though supporting the same underlying concept. There are few standards in terms of the capability provided, and no standard interfaces. Lock-in. Moving between suppliers is far from as simple as when changing traditional utilities.
Separation of concerns There are very good levels of separation between utility supply and its application. If I change my electricity supplier I am not constrained to only cook a certain brand of food as a consequence. Utility/cloud supplier’s stacks tend to be highly dependent. I may not be easily able to combine a capability in one layer of the stack from one supplier with a capability in a different layer from another supplier. Lock-in. Buying one capability from the utility/cloud supplier may limit you to buying associated capabilities only from the same supplier.
As well as lack of portability, there is lack of interoperability, either at a coarse or fined grained level.

So there is a breakdown of the analogy. Utility implies a fairly generic capability, but Cloud Computing is not always providing this. Applying the term to more specific capabilities might help people understand the business case, but they shouldn’t misunderstand the possible impact of their dependency upon them. However, this does not imply that cloud and utility computing are unable to deliver the promised benefits, providing the risks are assessed correctly.

To understand these risks better, and to help make the assessment, I have provided a framework for understanding the different models of utility and cloud computing and assessing them based on variety of factors.

See my reports Making Sense of Cloud Computing and also Service Portfolio Planning and Architecture for Cloud Services (both free on registration)

No comments:

Post a Comment