How wrong can IT projects go? Just ask the UK government. Its £41 million (about $61.4 million) Project Trawlerman, a defense intelligence system, was already out of date before it was delivered in 1996.
"This year the UK is going to spend 1% of its GDP (gross domestic product) on government information systems," says HP Labs' Richard Taylor. "And yet almost half of the new investment won't meet its initial requirements."
Understanding why this happens -- and finding ways to prevent it -- is the focus of Taylor's Performance Engineering team at HP Labs, Bristol, UK.
poor planning dooms projects
It's not just government IT projects that are prone to disaster. Consider that many of the abandoned UK systems were created by private multinational companies, and you'll see why the private sector doesn't have a much better record.
Such projects fail, Taylor believes, because they are poorly planned. As engineering projects go, IT projects are especially complex. Not only do they require technical expertise, they must accommodate the rapid evolution of technology and are therefore commonly dogged by changing project specifications.
Yet these realities are often forgotten when it comes to writing IT requirements into an organization's business plan. What's more, IT system planners often fail to effectively communicate with the software architects who design the IT systems that the business plan envisions.
The software architects in turn often fail to communicate with the people who put the software into operation. The result of any of these failings can be an IT system that is out of date almost as soon as it goes online.
new software standard
What's needed, argues Taylor, are ways of modeling such systems -- before they are built -- that allow the implications of any changes in the business, the IT architecture or its management to be explored before disaster strikes. But such sophisticated, dynamic models are extremely hard to create.
"What we've been doing," Taylor says, "is working on a set of modeling technologies that allow people to describe the business, translate everything that is relevant into terms that can be understood by the system architects and allow the architects to see the impact of the decisions they make on a system, and likewise with the operations people."
models that make sense
The result is that the different people involved in creating an IT system see models and results that make sense to them.
"The business plan people might see spreadsheets connected to a customer relationship management package that makes estimates," Taylor explains. "The system architects will see something that's far closer to the physical and logical reality of what they are putting together. The operations people may just see schedules."
HOLOS is based on an HP-developed modeling language called DEMOS 2000 that was architected by researcher Chris Tofts.
"With DEMOS," says Taylor, "we can take models that people have written in any of the classic ways of representing systems, do a lot of analysis within DEMOS and then we can export these models into other formats."
speaking the same language
DEMOS is to be released as an 'open source' public license, in the hope that it will become an industry standard for exchanging different types of models. Meanwhile, a HOLOS toolkit has recently completed a successful beta run. HP customers can look forward to seeing HOLOS used in HP information systems solutions such as HP's Process Manager.
"You will have much better integration," Taylor promises. "It essentially allows people to talk the same language without necessarily understanding the details of what other people are doing."
The value of this kind of modeling has already been proven within HP. HP's laser printer division in Boise, Idaho now has a formal modeling team looking at performance analysis, measurement and modeling. And partly as a result, HP was able to keep prices static and double performance on printers released during a month-long trial, which, says Taylor "was somewhat shocking for all of our competitors."
The next step is to take HOLOS out into the wider world. The group has plans to keep developing the software. They are currently looking at how to allow the continued application of the modeling tools throughout a system's lifecycle.
by Simon Firth