Heavy metal for the data age

April 30, 1999

Stephen Quigley celebrates the creation of the first working stored-program computer

"May 6th 1949. Machine in operation for first time. Printed a table of squares (0-99), time for programme 2 mins, 35 secs. Four tanks of battery 1 in operation."

Thus wrote Maurice Wilkes, director of the Mathematical Laboratory at the University of Cambridge, as the first entry in the log book of the Electronic Delay Storage Automatic Calculator (EDSAC), the world's first complete and fully operational electronic digital stored program computer.

It was not a particularly fast machine by later standards. It broke down frequently. Its peripheral devices caught fire occasionally. It occupied a whole room, 15ft by 15ft, that got so warm it provided enough heating for the coffee room below. But EDSAC marked the start of automatic computation in UK universities both as an academic discipline and as a service to students not involved in the machine's construction and operations.

To celebrate the 50th anniversary of its first operation, some 500 people whose lives had been influenced by it in some way recently gathered at Cambridge to shake hands with former tutors, mentors and colleagues including Dr Wilkes himself, and to relive the tale.

As early as 1936, the university recommended the establishment of a computing laboratory with the remit of providing a service for general use and of being a centre of computational techniques. It was founded the following year, as the mathematical laboratory, with J. E. Lennard-Jones, professor of theoretical chemistry, as its first director. It was requisitioned by the Ministry of Supply for the war years and returned to civilian use in 1945 with Dr Wilkes as acting director.

Wilkes, who at 86 is still active as a consultant to industry, became convinced of the feasibility of building an automatic machine after attending a course of lectures on electronic computing machines at the influential Moore school of electrical engineering at the University of Pennsylvania in Philadelphia in the summer of 1946. And it was there he got the idea of using mercury tanks for memory, the principal technical problem facing him.

He built up a small team of academics, research assistants and engineers and work on EDSAC started in October of that year. Construction of mechanical components was done in the laboratory's own instrument workshop, mercury tanks by the university's engineering department and electronics by a local radio manufacturer.

The machine used an ultrasonic delay storage system for main memory. Data was stored in the form of ultrasonic pulses circulating through tubes or "tanks" of mercury, generated, amplified and resynchronised by electric pulses applied to quartz crystals at each end of the tube. There were 32 tanks each capable of holding the equivalent of 16 10-decimal numbers. For electronic circuitry there were 3,500 valves installed in 15 racks.

Paper tape and teleprinter were the media for input/output operations. A library of subroutines was built up for numerical functions, including division for which EDSAC had no operation, and I/O operations. There was a basic form of microprogramming consisting of the hardwiring of "initial orders" on rotary telephone switches. These orders had to be transferred to the store before operations could begin. EDSAC performed 15,000 arithmetic operations per minute.

A computing service started in 1950 and soon became popular with research students in such fields as radioastronomy, structural chemistry and X-ray molecular biology. Demand was such that a priorities committee had to be established to approve applications for use. An operator would run programs by day. At night, authorised users could run the machine until it broke. Users had to fend for themselves in learning how to program.

The decade also saw the laboratory gain its independence as a university department. It built up its teaching and research staff and in 1953 it started its first accredited academic programme, a one-year diploma in numerical analysis and automatic computation. There were three successful diploma students in the first year and by 1960 the number had grown to 10 with seven PhD students.

EDSAC2 replaced EDSAC1 in 1958. It was the first full-scale microprogrammed machine, used ferrite cores for memory, with fast paper tape I/O, magnetic tapes and some read-only memory. It later supported graphical output devices, in particular a Calcomp plotter operating at 200 pen steps per second that remained a popular device into the 1970s.

EDSAC1 had been designed, in Wilkes's words, "with an eye to users' needs". EDSAC2 was extremely easy to use and it widened the user base in the university to some 200. It was more reliable and programming was made easier by the development of a high-level language, Autocode. It operated at about 10,000 instructions per second with eventually 88 K bytes of main memory. It was 40 times faster than EDSAC1.

EDSAC2 was much loved by its users and was the last mainframe users could actually touch. There was an emotional ceremony to mark its closing down in 1965, attended by users from all over the country. The last program was punched on black paper tape that caused the machine to play the Last Post.

The EDSAC family were single-user systems. EDSAC2 was replaced by the first multiple-access system, Titan. This was a cost reduced version of a Ferranti Atlas, the most powerful UK-made computer at the time.

The laboratory, with some of the work done jointly with Ferranti, developed much highly innovative system software that set many standards for systems of the future. The password protect feature in the main operating system, designed by Roger Needham, remains the standard in today's multi-user systems. To support multiple user files and interactive use, magnetic disks were installed. It ran at a speed of 0.25 million instructions per second with 0.75 megabyte of main memory and 128 megabytes of file storage. It was about 25 times faster than EDSAC2. In 1967, the laboratory made its computing service on the Titan available to anyone who wished to use it, 24 hours a day, seven days a week. The multi-user service was a great success and obtained recognition beyond the university. Users increased to 900. Three years later as other university departments became increasingly dependent on computing, the service was set up separately within the renamed computing laboratory, under its own director.

Titan, among whose lasting effects was its use, with the attachment of a DEC PDP 7, for the laboratory's initial work on computer-aided design, marked the end of the heroic age in computing at Cambridge. By the late 1960s, the laboratory recognised that for all the user-friendly features and advanced computational facilities of its in- house systems, it was in danger of becoming isolated from the mainstream. The scientific community needed compatible systems for program and data trading and IBM's System/360 architecture ,announced in 1964 and widely installed, had become the standard.

In 1970, Cambridge got the agreement of the computer board, the government body, funded to develop computing within UK universities, to purchase an IBM 370/165 and double the staffing of the service department. The laboratory had to develop a command language front end to maintain the quality of user friendliness. The laboratory also offered a service to other UK universities that required compatibility with IBM systems.

There were to be two successors to this mainframe system, the IBM 3081 in 1982 and the IBM 3084Q in 1989, before the laboratory discontinued use of a general purpose mainframe in 1994. During this period, "the golden age of the computing service", as David Hartley, its first director, calls it, the laboratory met the needs of thousands of users, ranging from complex problems of theoretical chemistry to email and typing a dissertation.

In the 1990s, the computing service's strategy belatedly altered to reflect changes brought about by the microprocessor as the computer ceased to be something precious whose use had to be optimised. And from mainframe support through distributed computing, the strategy moved to the provision of networking as the central service.

In 1992, at a cost of Pounds 3 million, the laboratory installed a backbone fibre-optic network linking 31 colleges and the university through 25 kilometres of cable to enable PCusers to access the service. Today, there are some 24,000 registered users of the service developed and maintained by a staff of 54, including engineers.

The teaching and research side of the laboratory has also flourished. Since EDSAC's first calculation, the practice of computing has led to many strands of research. Main lines in the 1990s include compilers, interpreters, distributed systems and communications, multimedia, graphics and animation, natural language processing and image processing. The laboratory has many collaborative links with industry and other universities in the UK and abroad. There are 29 teaching and research staff, 35 postdoctoral research fellows and assistants, and over 420 students studying for diplomas, Tripos and higher degrees.

As Robin Milner, head of the laboratory said: "With EDSAC a new form of human activity began."

to read this article.

Register to continue

Get a month's unlimited access to THE content online. Just register and complete your career summary.

Registration is free and only takes a moment. Once registered you can read a total of 3 articles each month, plus:

  • Sign up for the editor's highlights
  • Receive World University Rankings news first
  • Get job alerts, shortlist jobs and save job searches
  • Participate in reader discussions and post comments


Featured jobs

Lecturer in Economics

Royal Holloway, University Of London

Lecturer - Criminal Justice and Criminology

University Of The West Of Scotland

Lecturer in Aerospace Structures

Cranfield University

Accreditation and Planning Manager

University Of Greenwich