The main problem is that it's not a product: TECHNICAL ISSUES: Utility computing means different things to different people, writes Geoffrey Nairn: [SURVEYS EDITION]

The vision that IT should be as easy to use as an electric light is appealing. But at the moment utility computing is largely that-a vision - and there are many technical hurdles to prevent the vision becoming a reality. "Vendors, IT buyers and investors need to understand that while utility computing is the next big technology wave, it is not a product, but a vision," says Andrew Schroepfer, president of Tier 1 Research, a US-based market research firm.

Of course, the IT industry's visions often do not get realised. Nevertheless, vendors argue that utility computing is no pipe dream. Many components already exist. For example, load balancing, network attached storage and the clustering of computers and applications are used successfully in corporate data centres.

Even the idea of variable, demand-based pricing is not new. Two decades ago, when businesses mostly rented mainframes, IBM designed its systems with additional power that could be turned on at a later date in return for the customer paying a higher rent.

This raises the question whether utility, or on-demand computing, is as revolutionary as many vendors claim. Some users believe it is. "We view on-demand as a stake in the ground. It represents the beginning of the commoditisation of IT and we expect to see significant benefits," says Doug Rennie, managing director for transition and transformation management at JP Morgan Chase.

The bank plans to be one of the biggest customers for IBM's on- demand computing model through a massive seven-year Dollars 5bn outsourcing deal. IBM says it will use on-demand technologies to "dramatically change" the way it delivers IT to the bank. Instead of a fixed fee - as in traditional outsourcing - the price IBM charges will vary according to the use made of its services.

Mr Rennie believes this approach will deliver "significant economic benefits".

Even at IBM, which has done most to promote the idea, utility computing is very much a work in progress. Some IT users are sceptical that it will ever amount to anything concrete.

"My suspicion is that it (utility computing) is a little bit of a buzz phrase," says Mike Hampson, head of operations strategy at ABN Amro's wholesale banking division. The Dutch bank recently signed a Dollars 1.3bn outsourcing contract with EDS that includes a "significant degree of flexibility" regarding pricing. Unlike JP Morgan's utility computing deal, ABN Amro reckons it can get this flexibility from an outsourcing arrangement. There is no mention of utility computing.

Nevertheless, EDS also offers utility computing for those that want it. Its "agile infrastructure architecture" combines hardware, web services and transaction-based services to deliver EDS's particular interpretation of utility computing. Utility computing means different things depending on who you ask, and scope for confusion is rife.

Forrester Research believes that today there are really only three vendors that really count: Hewlett-Packard, IBM and Sun Microsystems. IBM has been making the most noise about utility computing, but Forrester claims HP has a small lead over IBM because it was first to ship products for server provisioning and storage virtualisation.

"Today IBM suffers from a complex, futuristic vision that has led it to neglect server provisioning," says Frank Gillett, an analyst at Forrester Research. HP recently signed a deal covering provisioning software from Opsware, a company set up by former Netscape co-founder Marc Andreessen. EDS also uses Opsware technology.

Opsware is one of many start-ups that have jumped on the utility computing bandwagon. Another is Ejasent, whose software helps measure the use of computing resources -a requirement for demand- based pricing. But organisations are reluctant to act as guinea pigs for these start-ups.

The big IT vendors are thus using the technologies of these start- ups to add substance to their often hazy visions of utility computing, either through marketing agreements or via acquisitions. For example, IBM recently acquired Think Dynamics, a Canadian company, whose software allows businesses dynamically to allocate computing power and storage according to changing demand.

Veritas, best know for storage software, recently bought Jareva, a pioneer of server provisioning software. Sun has made two important acquisitions to bolster its strategy for utility computing, called N1. Last year it acquired Terraspring, a competitor to Think Dynamics, and more recently it bought Pirus Networks, which specialises in storage virtualisation - another key ingredient of utility computing.

The deals aim to counter the criticism that Sun's N1 vision is aimed primarily at customers running Sun hardware and Solaris, its proprietary flavour of Unix. Yael Zheng from Sun Microsystems denies this is the case. "We intend to support heterogeneous environments," she says. Cingular Wireless, the US carrier, is using Terraspring software to manage a "heterogeneous" or mixed-vendor server environment.

Nevertheless, Sun's first N1 product, a provisioning server for blade computers, only supported Solaris when launched earlier this year. "The next release will support Linux blades," says Ms Zheng. HP, IBM and Sun claim their utility computing visions include support for other vendors' technologies. But analysts say customers should ask vendors tough questions about this issue.

Critics also doubt whether utility computing, using the products available today, really delivers a cheaper, more flexible way of managing IT. Forrester Research recently found that some customers had rejected buying Opsware software because of its high price. Instead, they planned to use cheaper proprietary tools from IBM and Microsoft.

James Hall, global head of technology and research at Accenture, argues vendors are today offering capacity-on-demand and financing schemes rather than real utility computing. True utility computing will require technology advances in many areas - grid computing, workload management and storage virtualisation, for example - and these technologies are still being developed. Integrating these constituent parts will be a major hurdle for even the biggest vendors.

Despite the plethora of announcements, utility computing is still mostly a vision. Nevertheless, not all is hype and some of the key technologies will no doubt prove their worth. Customers should buy these technologies, but they do not have to buy the whole utility computing vision.