The problem is that most business computers sit around doing nothing most of the time. The solution is to track them all down and relocate them to your giant datacentre, where you can standardise, integrate and consolidate all your servers, thus benefiting from HP's Utility Data Centre.
Listening to Tomlin gave me a bad flashback. Suddenly I was 20 years younger, listening to IBM extolling the virtues of the mainframe. And dark, stultifying clouds started to gather on the horizonÉ
Of course, there has never been anything wrong with building large datacentres: they are often the cheapest and most efficient way to do things. I have never confused "the death of the mainframe" with "the death of the datacentre".
The problem with the traditional IBM liquid-cooled mainframe was that it was a hugely expensive monster and it was incompatible with almost everything else in the IT business. That included IBM's three or four ranges of minicomputers, which were also incompatible with one another. Thus, although it cost a small fortune to clamber into IBM's mainframe lobster pot, it cost even more to get out.
But although the business case for a Utility Data Centre may be the same as it was for the mainframe, Tomlin said, "At least HP is doing it with open systems. We are building things out of very cheap components, and people get incredible value for money."
And you get a choice of operating system: you can load Windows or Linux, or whatever else you like, and configure the whole system remotely.
Clearly there are benefits to building mainframe-class systems using PC parts, where prices are reduced by competition and economies of scale. And in theory, having a Utility Data Centre could encourage even more creative experiments than the availability of cheap Dells. But I suspect it will not: the datacentre mindset could turn out to be just as stultifying today as the mainframe mindset was in the past.
Jack Schofield is computer editor of the Guardian
This was first published in September 2003