Eating out is the next big thing. Rather than buying all of the ingredients for the meal, the kitchen, the pots and pans, the utensils and then cooking it, on-demand meals are the answer.

How can I have convenience without any capital investment is the question. Utility computing is the answer.

You can eat when you want, you can eat as much as you want and you can choose what you have from a menue.

So what is this Utility computing thing?

It sounds like a good idea. It meets the business requirement. It is flexible. It is more efficient than the current model.

Currently an organisation requires sufficient computing infrastructure to handle the peaks as well as the normal loads. The new e-business model makes it more difficult to predict the demand. The consequence of this is a requirement to have sufficient resources available for an unknown demand. This can be quite expensive.

The theory is that by employing on-demand computing the peaks and troughs are handled by the load-balancing of aggregated resources. This is not going to solve all of the problems. If it is to work like a true utility then it will require unambiguous standards, ubiquitous high-bandwidth networking and a scalable architecture.

On top of this will be the security perspective.

These features are not available now and it will be several years before they are mature enough to be a true utility. Until then the offerings will be through traditional bureaux services, ASP and Grid computing.

Feature Story

"Pay as you go" computing is coming

It is wrong to think that the on-demand computing utility model will be significantly cheaper than cooking it yourself. To get the benefits of the utility model, businesses must decide what they will be doing in-house and what they want to do through utility. Once this has been done then sensible decisions can be made as the utility offerings become available.

To understand the model, I recommend that companies create geographical federated services across their businesses e.g. enterprise Grid computing. To revisit the systems architecture with a single objective to de-couple the applications from the storage. A policy should be implemented to stop the purchase of directly attached storage (DAS) in favour of network attached storage (NAS).

Wherever possible server clustering should be implemented for servers (Scale out). This enables the easy introduction or removal of server capacity as required. Future operating systems will make this into a non-intrusive dynamic capability. Databases should be put on servers with dynamic domain capabilities that allow the easy introduction / release of computing resources - CPU, memeory, cache and I/O - (Scale up)

The businesses should then review their business portfolio and decide which parts of the business will be subject to high growth or contraction. These are the areas that will require a highly responsive computing provision and could benefit from utility computing. In todays business world under- and over-provisioning is no longer an option.