Background

ICL was created in 1968 as a result of a merger between ICT (International Computers and Tabulators) and English Electric Computers; it was to be the mainstay of the U.K. computer industry. The merger brought together the major U.K. suppliers of commercial computer systems at that time; they had between them some 50 percent of the U.K. mainframe market.

The merger was encouraged by the U.K. Labor government, which took office in 1964, and believed that the indigenous U.K. computer industry was on the point of extinction (1 ). This recognized the realities of the investment costs of competing in the international computer market. ICT had a highly successful range of computers, the 1900 series, and some 2000 of these systems had been sold. Its architecture was based on 6 bit characters and was viewed by some as dated. English Electric was marketing a range of computers known as the System 4 range, which was IBM 360 compatible and was made in the U.K. under license from the RCA Corporation. Only 200 systems had been sold, but the architecture was competitive and used 8 bit bytes. These two ranges of computers were incompatible with each other.

A decision was taken to develop a new range of computers as a result of a study of various options. The technical case for the choice was based on competitive comparisons and research and development work worldwide. It drew strongly on work being carried out at Manchester University, U.K., where the MU5 was the latest in a line of developments which had led to other major advances in their time, such as the Atlas computer. It also drew on research at the Massachusetts Institute of Technology, U.S., and the University of Eindhoven, Netherlands, where Dijkstra had been developing concepts for structuring operating systems (2).

The marketing decision carried risks. The justifications given for the decision included technical and internal political factors and not simply commercial reasons. It was thought that a new ICL product would help to build the new corporate identity, and this case was advanced as a break with the past. The strategy was high risk because of the usual need for customers to be able to transfer their programs onto a new machine. Lack of forward compatibility removed the competitive advantage normally enjoyed by a supplier in terms of its existing customers. To try to sell customers machines which were incompatible with their present ones increased the risk of losing them to competitors. A customer who had been tied to ICL for years by the proprietary design of the software would have a free choice of supplier when considering a move to the "new range." However, on the other hand, there were the technical issues of the two existing incompatible ranges and the outdated design of the majority of the company's systems.

The development of the "new range" (the 2900 series as it became known) took five years from the initial design documents to the early prototype systems. Some 200 programmers worked on the operating system alone, and this cost 1,000 staff years of development effort before the first version was released for limited field use in 1974/75. The cost of this development had been budgeted for, but the work took longer than expected, and it was toward the end of this period that the managing director of ICL, Geoff Cross, turned his attention to the development side of the company.

Geoff Cross was recruited from Univac in the early 1970s to rescue ICL, which was losing money in a stagnant market; there was also increasing competition from minicomputer manufacturers such as Digital Equipment Corporation and Data General. His first priority was to address the sales situation, which he did with considerable success, increasing sales volume by sharpening up the sales force, introducing new products such as the 2903/4 minicomputers, and developing new ways of selling, such as the creation of customer centers offering bureau services.

By 1974, time was running out for the older models as the customer base was being won over to the new 2900 systems. The sales volume of the old systems was declining, and the first 2900 systems were due to be shipped at the beginning of 1975. The business plan required that they be delivered on time m to maintain the revenue stream. The director of the systems programming division (SPD) was having difficulty holding development work to deadlines.

CADES Approach to Software Development

For some years, the SPD had used project management techniques, such as PERT, to help monitor and control development work. From the beginning, the VME/B operating system development work was supported by the use of a database system, CADES (computer-aided design and evaluation system), and modular design techniques. CADES was used as both a management tool and a design aid.

A structured approach to software development was becoming fashionable at this time, after years of anarchy when it was unusual for a computer programmer to stick to deadlines. Folklore indicated that any program would take twice as long to produce as planned, regardless of whatever forecast was made. The computer industry's version of Parkinson's Law was that work expanded to fill twice the time available! In an attempt to bring more professionalism to software development, engineering techniques were adopted. This is one facet of what is now referred to as "software engineering."

As in network analysis, structured software design and modular programming techniques require a problem to be broken down into a series of small manageable tasks. A module of code is then written to perform each task. A sound design will identify how many modules are required and how they interrelate. It is possible to design each task to require a similar amount of code. On the basis of experience, it is possible to estimate how much testing is required to get a module working. This makes it possible to estimate fairly accurately how much work is required to complete the overall job and to monitor progress. CADES was used to support this management approach.

A team of 200 systems staff was involved in the development of VME/B for a period of five years through to the early releases. This amounted to a huge project that comprised in excess of 1,000 staff years of effort. This was a massive investment, which cost some £50M in today's terms, that had to be managed.

A central design team of twenty people was responsible for the development of the operating system, which was divided into subsystems, as shown in Figure 1. Each designer had responsibility for one subsystem. The designer monitored the production of the subsystem under the direction of the chief designer. The production was undertaken by programmers in teams of ten to twenty, under a line manager. Each of these was responsible for writing the modules of a subsystem, and one member of the design team was responsible for the work of each programming team. A central support team of twenty was responsible for running the CADEs database, which was an overhead of 10 percent of development costs. f

As modules were written, their specifications were coded and entered into the CADES database, which performed certain checks, eliminating some syntax and logic errors. Software interfaces were checked for correctness and access permission. Procedure calls and data definitions were generated, thus reducing the risk of errors as well as capturing design and usage data.

Project Management Made Easy

Project Management Made Easy

What you need to know about… Project Management Made Easy! Project management consists of more than just a large building project and can encompass small projects as well. No matter what the size of your project, you need to have some sort of project management. How you manage your project has everything to do with its outcome.

Get My Free Ebook


Post a comment