There is a strong link between on-demand or utility computing - which provides IT resources that can quickly be adjusted to meet changing demands - and grid computing, which connects multiple computers so capacity can be pooled and accessed more easily.

However, in recent weeks experts have warned that utility computing faces some obstacles, including political issues associated with the introduction of the technology.

Research company Gartner said that many government departments and managers will not allow their IT resources to be shared. And the same difficulty is likely to affect many businesses.

It is not that managers are shortsighted - most would happily contribute to the pool of utility or grid resources, but only if they are sure that this will not jeopardise their applications. Clearly utility computing will be undermined if managers do not allow spare capacity to be shared.

The case for sharing is strong, however. Intel recently said that the average utilisation of CPUs in a typical Intel-based server is not 30 percent as commonly quoted, but only a paltry one percent.

I find this figure stunning. It does not mean that firms will not need new servers for many years, though. This is because such systems may operate at an average utilisation of one percent, but they probably peak at 100 percent utilisation for short periods.

It is clear that there is a colossal amount of spare computing power available to some firms, if systems can be set up to exploit it. This provides justification for utility and grid technologies, but applications may need to be upgraded before they will work well in the new distributed environments.

Setting the political issues aside, standards for utility computing have room for improvement. Tony Storey, IBM fellow and chief software architect, said recently, "On-demand services can be seen as document-oriented distributed services built on standards such as Soap, WSDL, UDDI etc, but [there is more work to be done]. The key now is the development of higher-level applications services. These mainly fall into the area of transaction management rather than just message passing."

In other words, for many utility systems, the applications and management tools will have to become better integrated, to ensure that tasks are completed in all their stages, and in the right order. Otherwise, service provision will fail.

Guy Bunker, chief scientist at Veritas, speaking at a recent Storage Network Industry Association briefing, said the time is now ripe for the rebirth of programming for the VMS operating system - because the potential of utility and grid systems will only be fulfilled if there is a new generation of distributed-applications developers.

Currently, applications are not written to exploit the capabilities of utility and grid systems. As a result, applications are often greedy resource-hoggers. In future they will have to be designed to share resources and work together, but I don't see many software vendors working on these lines yet.