Information Age: News, analysis & insight for IT & business leaders

Organisations have already started experimenting with such technologies, indicating that while they do not want to throw themselves rashly into this new world, they are only too aware that something is broken with today's architectures that needs fixing.

If any one word sums up this shortcoming with current architectures, it is 'inflexible'. The large 'farms' of servers and storage devices that organisations have built up over the past five years are in many cases vastly under-utilised. IT executives cite cases of servers, even when running at peak loads, using only 50% of their CPU and memory capacity; other systems are lucky if they tax 20% of their power.

Inflexibility is the very reason for that over-capacity. Many servers today are application-specific. As such, they are unavailable to other applications which could use their spare capacity.

What organisations want to do is to bring that all together, to pool those resources and share the applications requirements across them. And the message organisations are sending vendors is clear: 'provide us with the means to get to that more efficient exploitation of resources'.

Getting there is a multi-stage journey. Even by 2008, when IT industry watcher IDC says organisations in Western Europe will be spending $1.8 billion a year on grid computing servers, that spending will only represent around 10% to 12% of their total server purchases.

As that suggests, grids will not necessarily be built from new servers, but by organisations making better use of their existing systems. Nathaniel Martinez, an analyst at IDC's server group, suggests that organisations will adopt technologies that support the new model on a piecemeal basis, moving unit by unit, application by application. "Adoption of server computing grids is slowly moving beyond the early stages of concept proofing and testing," he says.

Much of that early adoption has been in compute-intensive industries such as pharmaceuticals or financial services. But, over the next few years, grid computing will become pervasive in commercial data centres, says Martinez.

Barriers to adoption still have to come down. Departmental 'ownership' of servers or other resources has to give way to an acceptance that IT can be delivered as an on-tap, utility service, available in whatever quantity the business unit requires - no matter whether the power to do so is coming from internal or external resources.

Applications also need to change their structure. Part of the issue of flexibility is that organisations have not been able to adapt their applications to match their rapidly changing business processes. The service-oriented architecture is designed to tackle that by breaking applications into small, discrete modules and assembling and re-assembling application services to support changing requirements.

While it may take some time for organisations to bring their applications into an SOA environment, the software for creating a virtualised view of the pooled resources and for provisioning resources to meet application needs is largely there.

Virtualisation products from companies such as VMware provide a layer of abstraction between the computing, storage and networking hardware, and the software that runs on it. By moving applications to where they run most efficiently, VMware claims customers can push utilisation rates on Intel-based servers between 60% and 80%, up from 5% to 15% today. It also maintains that application movement can be dynamic, as provisioning times for new applications can be measured in tens of seconds rather than the days it would normally take to rehost an application on a new server.

There is little disagreement among IT executives of the desirability of a move to that kind of model, where application resources are provided when they are needed and in a sufficient - and efficient - quantity. For now, many are working on what some regard as a prerequisites for utility IT, the consolidation of distributed resources into fewer, larger data centres; many of the same organisations are packing out these data centres with blade servers with a view to creating a grid architecture; and almost all are building an SOA.

The upshot will be profound for the structure of the IT industry. In the short-term, it simply means the more efficient use of technology. But few doubt that this is the first step on the road to utility computing, where commodity IT capacity will be generated and delivered by external IT powerstations or collaborative grids rather than by stand-alone IT departments.

Whatever the shape and however long that takes, IT is destined to progressively lose its reputation for inflexibility.