In the 1980s, companies in the UK began to offer metered water to households instead of flat rate charges based, indirectly, on property values. Many homes that switched over to metered water found that their bills fell. However, for some water costs rose.
Families with above average water consumption - perhaps because they had several messy children or extra garden sprinklers - lost out because they did not realise their consumption was so high. After all, who knows how much they will pay for a metered service, until someone meters it?
Potentially, companies face the same problems as they move to utility computing. IT companies that promote this pricing model stress the financial benefits to their customers of paying for what they use. But companies can be sure of savings only if they have a very clear idea of their IT costs.
"It is scary how little organisations know about what they are paying for their IT, especially large organisations," says Alastair McAulay, of the global IT consulting group at PA Consulting. "Very few companies can go to a single place and find out what they spend." Without this, he suggests, it is hard to know how much a business can save.
Combine this lack of knowledge with the absence of standardised pricing among vendors of utility computing, and few third-party tools to calculate the computing capacity needed for a particular business process, and it is easy to see that CIOs are taking a leap into the unknown.
IT vendors and users alike agree that computing resources are often very poorly utilised. When mainframe computers dominated IT systems, utilisation reached 80 per cent. With Unix, Windows and other "distributed" computer systems, utilisation rates can be as low as 20 per cent. Businesses are unlikely to increase their IT budgets until this is addressed.
Utility computing promises a solution, but unless businesses manage it well, they might not gain benefits. "Most people who buy into utility computing do so because they believe it will allow them to achieve better utilisation of technology," says James Hall, global managing partner for technology and resources at Accenture, the IT consultancy. "But the different players have different views of what utility computing means."
In some cases, businesses will be paying for IT as a utility-type service; in others they will pay for hardware, but in much smaller increments than they do today.
Developing standard measures for common elements of computing power, such as a processor second or hour or a gigabyte of storage, should be relatively easy. Scientists at Hewlett-Packard recently did just this, creating a standard unit for IT capacity, known internally as a "computon".
But HP admits that there is work to be done before the computon is as easy to use as a kilowatt hour of electricity. And, unlike commodities, there are no guarantees that HP's computon would be comparable with a unit of computing power from Sun Microsystems or IBM. "In utilities there is a standard measure of capacity, but that assumes a regulated definition of performance," says Mr Hall.
Nor does a measure of computing power answer the needs of managers, who are more interested in the cost of business processes than of hardware. Many utility computing contracts use transaction measures, such as a database entry or accepting an e-commerce order, rather than capacity. This makes comparisons harder still.
This leaves CIOs with a choice. They can buy raw capacity from an IT company, and work out internally how many processor cycles or how much storage they need. Or they can pay by the transaction and rely on the vendor's calculations.
"There is no accepted standard for a transaction. There will be some things you can commoditise, such as storage or perhaps an Oracle database look-up," says Mr McAulay, at PA. "But to bill by the transaction, you have to know what that transaction is.
"Either route has risks. Competition between vendors means that few companies should end up paying more for their computing power under a utility model, but there is a danger that they will not reap all the potential savings unless they are clear about their costs.
It would be all too easy for vendors to offer an attractive saving of perhaps 20 per cent or more to clients, knowing full well that they could save twice that through economies of scale and the latest IT management tools.
Vendors, for their part, argue that rather than wait for the dream of true utility pricing, IT buyers should look at each deal on its own merits. "Even if we had all the hardware elements on a fungible, universal basis that would still not cover staff and other overhead costs," says John Lutz, vice-president for e-business on demand at IBM.
According to Nick van der Zweep, director of utility computing at HP's enterprise systems group, a lack of standard pricing is not unique to pay per use pricing models.
"We have different pricing structures in the industry even without utility computing," he says. "Whether you look at buying a computer system from Sun or IBM or HP the customer needs to look at the return on investment and the total cost of ownership. Customers are pretty savvy today and ask us for the three-year TCO.
"Highly detailed pricing models might not be what businesses want, suggests Bill Mooz, Sun's senior director of utility computing. "Business managers need granular cost data that they can use to drive the right behaviour. There have been complaints that some utility computing schemes are too complex to be manageable, but I don't want to know every time someone sends a PowerPoint file. I want to know the overall levels of usage, and if I see deviance from those norms I want automated tools that alert me to that," he says.
Some businesses will choose utility computing because of its flexibility, even if is not the cheapest option. IT managers might be happy to pay a premium to buy standby capacity when they buy new systems. Some vendors offer price caps so the total cost over three years is no higher than that of buying the larger machine from the outset.
"Some clients are pushing for this way of paying for IT, especially if they have asymmetrical patterns of usage and find it expensive to buy what they need for the peaks," says Mr Hall.
