Computer maintenance is a bugbear. The answer may be automated programming. Simon Vail assesses it.
It is a racing certainty that tomorrow's winning companies will own flexible computer systems. Today's computing realities are rather different. Many businesses are stuck with mainframe programs written 20 years ago, which are now wildly out of step with the realities of the global marketplace. They need armies of programmers to nurse their whims - the Gartner Group IT consultancy estimates there are over 2.7 million commercial programmers around the world who spend no less than 80% of their time maintaining existing systems, rather than developing new programs. It is a huge waste of resources.
Vendors of Computer Aided Software Engineering (CASE) dangle the promise of automated programming in front of companies, lumbered with massive systems. They claim to be able to modernise the 100 billion lines of COBOL code running on the world's mainframes. Programmers will disappear, or become analysts producing flawless computer code automatically. More importantly, they say that CASE overcomes computing's Achilles heel: neither the boardroom nor the information technology department talk each other's language. But the CASE market is fighting to shake off a poor reputation. It is renowned for swallowing as much money as it claims to save, demanding hours of management time and the possible restructuring of business practices. It is too daunting and too expensive for most companies.
CASE critics says its benefits are difficult to quantify. It is an emerging technology and has not been in existence long enough for anyone to prove its systems cost less to maintain than traditional ones. Julian Hewett, co-author of a report on CASE by the Ovum consultancy, says that there will be negligible productivity gains in the analysis and design of programs using CASE. The benefits come further down the line: the process of coding improves by three to one, and the most significant benefit is in maintenance, where improved analysis brings a payback of between 50 and 75%.
The technology is not for the faint-hearted. Although the phrase spans a multitude of productivity software tools, the all-embracing CASE market is known as Integrated CASE, or I-CASE. It combines tools for program analysis (the front end) with tools for code generation (the back end). Integrated CASE requires serious money, or as the vendors would have it, a 'strategic investment'.
I-CASE vendor James Martin Associates, recently bought up by computer company Texas Instruments, persuades blue chip companies running large IBM mainframes to part with between £250,000 and £500,000 for its Information Engineering Facility (IEF). Entry level workstation- based systems cost less but are not as comprehensive. Such a sum will buy the software tools needed to design and code an order entry system for a medium-sized company employing 1,000 people, says managing director David Fairbairn. He admits that compared with traditional programming techniques, savings in the first year would not be enormous, but he would hope to develop the system in a third of the time - six months instead of 18 months. The high upfront costs of buying the software are offset against more rapid development of systems, since the main business activity need be analysed only once.
Time-saving of this order is achieved by the replacement of traditional paper and pencil program analysis with a rigorous, screen-based system that records business activities and future trends as a series of single nouns and verbs, illustrated by diagrams of interlinked boxes. All information is held in a central library or repository, to which all program designers have access. A simple statement for a knitwear manufacturer like 'We sell socks' is broken down into careful analysis of who 'we' are, how socks' are made, and what 'sell' means. Once the activities of the company have been broken down into logical statements, they are more easily translated into a programming language like COBOL and turned into a marketing system, for example.
'What we are trying to do is find a high degree of precision that is maintained throughout program design, which will lead to higher quality programs that are bug-free. If you can define with absolute precision every step in a process that is necessary to produce a Ford car, then the process can be automated. The same potential exists for system development,' says Fairbairn.
Rolls-Royce has used IEF since 1987. The company began to investigate CASE when it acknowledged that business aims and computing policy had drifted apart. 'We realised that we were not translating the strategic vision of the business into practical and cost-effective systems,' says system director Wal Budzynski. 'Our traditional systems were reflecting departmental needs, rather than the total requirements of the business.'
The IEF toolkit forces Rolls-Royce to examine the nature of the business task as closely is the technical requirements of programs used to help manufacture jet engines. The company has used IEF to develop complex systems for bill of materials, sourcing, an integrated engineering plant and a company-wide logistics operation that will reduce 300 separate systems to 25 company-wide systems. Budzynski estimates that CASE has given the company between 100 and 200% gains in productivity and a great reduction in defective programming.
Other successful CASE users are similarly enthusiastic. British Gas is an experienced user and has spun off a subsidiary called Information Architects to market its own CASE tool. The potential of CASE is said to be enormous. Intersolv, another I-CASE vendor, predicts that 60% of major corporations will select a CASE system in the next two to three years.
The company also points out that whereas in 1988, 60% of program development used to take place on main-frames, 45% has moved to PCs linked by local area networks. PCs suit CASE since they offer very good graphics. Programs that are easy to use sell well. 'While you can't expect the users to do all the tasks involved in building a systems, the more heavily involved they are in the front end, the more likely the back end will be what they want,' says Andersen Consulting's CASE partner Rob Baldock.
He is convinced that effective program development takes place when both systems departments and users are committed to a new system. National Power has taken Andersen's advice to heart, and spent over a £1 million on Andersen's I-CASE tool called Foundation. The electricity company has embarked on a £300-million program of system development, following the break-up ofthe Central Electricity Generating Board. Tony Davis, National Power's head of IT quality, says he inherited a mix-'n-match series of mainframes, minis, PCs and incompatible operating systems from the CEGB. 'The old set-up was full of bits and pieces. It was nowhere near good enough for our commercial environment.'
The company has adopted CASE technology because, it says, the management of such an investment is easier to control, staff need to be trained on only one set of applications software and all program design and analysis information is held in a common format. Staff, for example, are not called 'personnel' in separate databases. More important, says Davis, was a strategic decision to develop systems that were integrated seamlessly with the business activities of the company.
'We think it is so important that every business application is sponsored by a director or senior manager within the main part of the company. Every development has a steering committee that is chaired by senior manager who is not directly involved in IT,' Davis adds.
National Power is halving the staff it inherited from the CEGB, reducing the headcount from 4,000 to 2,000 over the next 18 months. CASE-inspired computer systems, standardised round one Amdahl mainframes, Digital Vaxes and Compaq PCs, have devolved day-to-day decision making to 40 power stations round the country and made the remaining staff more productive.
The National Power experience takes CASE closer to the development of 'end-user computing', where non-IT managers write their own programs, understanding the system enough to tinker with it. 'Users accept systems more readily when they have seen a prototype of how the system will look, and can make suggestions about screen design which the IT department understands,' says Tony Blench, LBMS product marketing manager for the popular I-CASE tool called Systems Engineer.