Tom Kilburn: A tale of five computers
One of the pre-eminent figures in the early history of computer design was Tom Kilburn. Over the course
of some thirty years, he made significant contributions to the development of five important computers.
Although a natural team leader possessed of a somewhat dominating personality, who inspired in those
who worked closely with him great loyalty and affection, Kilburn was, on casual acquaintance, a self-
contained man who chose his words with care. As F.C. Williams put it “What you must always remember
is that Tom is a Yorkshireman.”
1
Early Days
Tom Kilburn was born on the 11
th
August 1921, near Dewsbury in West Yorkshire, England. His father,
John William Kilburn, was a statistical clerk who rose to become a company secretary.
1
Tom had a
somewhat specialized education at Wheelwright Grammar School having been permitted by his headmaster
to study almost nothing else from around the age of 14. It was hardly surprising therefore he emerged from
school as something of a mathematical specialist. In 1940, Kilburn went up to Sidney Sussex College,
Cambridge, with State, Dewsbury Major, and Minor Open Scholarships. Wartime courses at Cambridge
were somewhat truncated and in 1942, Kilburn graduated with First Class Honors in Part I of the
Mathematical Tripos and in the preliminary examination for Part II.
During the Second World War, many Cambridge mathematics dons were absent from the university
serving at Bletchley Park and elsewhere. In spite of this, there remained a lively mathematical community
in which Kilburn played a full part part. As the Sidney Sussex college representative in the New
Pythagoreans (a subgroup of the Cambridge University Mathematical Society), Kilburn almost certainly
came into contact with a number of people who later went on to play a part in the development of
computing. Geoff Tootill and Gordon Welchman were, like Kilburn, officers of the New Pythagoreans.
2
Speakers to the student society included future Bletchley Park code breakers M.H.A. Newman,
3,4
K.J. Le
Couteur, and William (Bill) T. Tutte. However, Kilburn would not have been likely to have come into
contact with Alan Turing, who spent most of the years 1937-38 at Princeton based in the world-leading
research group in mathematical logic headed by Alonzo Church before taking up his position at for
Bletchley Park in 1939. It is also unlikely that Kilburn read “On Computable Numbers” as an
undergraduate because his mathematical taste was more applied than pure:
[P]ure mathematics seemed extremely abstract. I was the sort of person who was always prepared to accept
that two and two are four, whereas I’d spent the first term at Cambridge in one of Newman’s lectures proving
that this was so. Whilst it was all very interesting—I mean one could appreciate the beauty of it—it left me
rather cold. At the end of it, you didn’t seem to be much further forward…
5
War Service
At some point during his final year at Cambridge, Kilburn attended a talk given by C. P. Snow,
6
who
was visiting universities recruiting people for unspecified war work. By his own account, Kilburn had some
fairly settled ideas about what he wanted to do for the war effort:
It seems silly but if I could have joined the RAF as a pilot, I would have done that, but I was relegated to
navigator or some such, and that was not quite so appealing. It sounds egotistical but I like to lead. I like to be in
charge and I didn’t fancy the idea of being driven and crashed by some other character. I wanted to do my own
driving and crashing. It’s on these sorts of whims that life is founded—it’s not through any profound thought is
it? You take advantage of what’s there at the time.
5
Having ruled out service in the RAF, Kilburn enrolled on a number of short courses in electronics. He
then took a six-week City and Guilds course on electricity, magnetism, and electronics. After about a week-
long vacation, he was assigned to the Telecommunications Research Establishment (TRE), Malvern, where
he joined Group 19, led by F.C. Williams. He was not, however, greeted with unbridled enthusiasm.
Williams had requested an extra person to join his team, and Kilburn was the person they sent. The other
members of Williams’ group were all around 30 years old with an average of 10 years practical experience
in electronics. By contrast, Kilburn was 21 years old and, prior to being drafted, had not the least interest in
electronics or electronic equipment of any kind. Williams, whose group was responsible for designing and
debugging electronic circuitry and for solving problems encountered by other groups, made no attempt to
hide his disappointment at being offered someone so inexperienced. Kilburn later recalled:
[I]n effect he said “Oh God, you don’t know anything?” and I said “No.” That was the sort of relationship at
the start. But of course by the time we left Malvern—that was four years later—the relationship was quite
different.
5
The Manchester Baby
By the end of the war Kilburn was well settled domestically, having married Irene Marsden in 1943, and
professionally, he had, despite such an inauspicious beginning, become an important member of Williams’
team. Kilburn had achieved the rank of acting scientific officer. In 1946, Williams left TRE to take up the
Edward Stocks Massey Chair of Electro-Technics at the Victoria University of Manchester. It was
Williams’ intention to continue his work on the development of the cathode ray tube (CRT) memory, and
he arranged for Kilburn to work with him at Manchester under secondment from TRE. By the end of 1947,
Williams and Kilburn had developed a CRT that could store patterns over long periods. But as Kilburn put
it, “the only way to test whether the cathode ray tube system would work in a computer was, in fact, to
build a computer.”
7
The story of precisely how the Manchester Baby was conceived, funded, and developed as well as the
roles played by various actors in the project is somewhat complicated.
8,9
The dominant historical narrative
has come univocally from the engineering tradition and has generally paid little attention to the contribution
made by people like M.H.A. Newman and P.M.S. Blackett.
10
What is not in doubt is that the machine itself was the first working example of a digital electronic
stored program computer. However, its significance for the historian of computing is not that it was an
iconic first but that it provided the foundation that Manchester used to build itself into a leading center for
the emerging computer science field.
Mark I and Mercury
Kilburn’s original intention had been to return to TRE after the Manchester Baby was completed, but the
success of the Baby was such that the Ministry of Supply quickly awarded a contract to Ferranti to design
and build a full-scale commercial computer to Williams’ specification. An important initial step was to
construct a prototype machine,
11
the Manchester Mark I, which was to be produced at the university. It was
clear to all that Kilburn was vital to the new project, and Williams, partly with the inducement of a
lecturing post, was able to persuade him to remain at the university to work on the prototype.
By the autumn of 1949, the Manchester Mark I, now with backup drum store, was complete, and ran
continuously thereafter for almost a year. Between 1951 and 1957 around nine of the Ferranti production
versions of the Mark I machines were sold.
Over the three years since the Baby was completed, two important shifts of responsibility had taken
place. First, Max Newman came to the conclusion having provided the initial impetus and leadership in the
development of the Baby, further computer development required engineers rather than mathematicians to
be in the driving seat. Newman’s withdrawal left Williams in sole charge, but Williams, who never had
much interest in computing as such, fairly quickly passed effective control of further developments to
Kilburn.
In 1951, once again following a process of incremental development, Kilburn began working toward a
Mark II computer that was known as the megacycle machine, or Meg. It replaced the Mark I valve diodes
with solid-state versions and offered a tenfold increase in clock rate together with greatly improved
reliability and floating-point operation. The serial CRT memory, which was already running at a near
optimal rate in the Mark I, threatened to act as a performance bottleneck for the Meg. Kilburn’s solution
was to design a 10-bit parallel CRT memory.
Meg first operated successfully in the summer of 1954, and Ferranti developed a commercial version of
Meg under the name Mercury. Clients included the Meteorological Office, the Norwegian Defense
Research Establishment, and Manchester University. In all, 19 Mercury computers were sold, of which 6
were purchased by overseas customers.
Transistor Computers
Kilburn led a Manchester design team consisting of Dai Edwards and Tommy Thomas, who
concentrated on the Meg, together with Dick Grimsdale, and Douglas Webb, who were simultaneously
working on what was originally a research project looking into developing the smallest possible economic
computer.
12
It was soon clear that a great deal of valuable experience could be gained by using transistors
to build the machine. Two prototype transistor computers were commissioned, both of which made use of a
pseudo two-address instruction format and permitted optimum programming. The 48-bit machine produced
in November 1953, which is widely acknowledged to have been the world’s first operational transistor
computer, had 550 diodes and 92 point-contact transistors, was manufactured by STC. A somewhat
enhanced version of the transistor computer was completed in April 1955, this time boasting some 1300
diodes and 200 point-contact transistors.
13
In 1956, the Metropolitan Vickers Electrical Company, adapted the design of the experimental transistor
computer to allow the use of junction transistors, and manufactured six transistor machines, mainly for
internal use, under the name Metrovick 950. From the perspective of Kilburn and his team, the most
important aspect of the Transistor computer was the early experience it gave them in transistor circuit
techniques.
14
Muse and Atlas
Kilburn’s plan with the Muse (microsecond) project was to develop a really large fast machine that
would make full use of both existing and emerging technology. He succeeded admirably; Muse used
“multiprogramming, job scheduling, spooling, interrupts, pipelining, interleaved storage, autonomous
transfer units, virtual storage and paging—though none of these techniques had been invented when the
project started in 1956.”
13
Muse can be considered comparable in scope and ambition to the IBM Stretch and Univac LARC
projects, but Kilburn was under no illusion that the Department of Electrical Engineering had sufficient
resources available to complete, without assistance, a project of this scale and complexity. His initial
attempts to elicit Ferranti or the government’s support for the Muse proposal were unsuccessful, so the
decision was taken to proceed instead with a trimmed-down version of the original scheme. However, in
January 1959, Ferranti, with £300,000 in backing from the National Research Development Corporation
(NRDC), decided to participate in the project, now renamed Atlas.
14
Among the innovations introduced in the Atlas was a scheme that let programmers treat drum stores as
if they were core storage. An innovative program called the supervisor managed drum transfers. The one-
level-store concept—the idea of a fast and a slow store appearing as a single fast store—was an important
precursor to virtual memory.
Kilburn was not only responsible for the management the project but also played a part in the circuit
design, including work on an adder with a fast carry path.
1
In general though, he relied substantially on the
experienced teams that he had established over a number of years. Three Atlas systems were eventually
built and installed at the Universities of Manchester and London and at the Rutherford Laboratory.
In 1960, Kilburn was appointed a professor of computer engineering in the Department of Electrical
Engineering.
Department of Computer Science
In addition to his considerable contribution to early computer design, Kilburn also did much to establish
computing as an academic discipline in the UK Higher Education curriculum. Starting in 1963, he spent
several years establishing and organizing a new Department of Computer Science, the first of its kind in the
UK. The intention was to provide a natural home for computer research, a sound base for future projects,
and undergraduate courses in computer science. Kilburn, now translated into a professor of computer
science, was the inaugural head of department, and had 12 academic staff under him. Directly reflecting
Kilburn’s personal strengths and his professional experience, Manchester placed more emphasis on
hardware than many of the other computer science departments that followed it, most of which sprang from
a mathematical lineage rather from engineering. Kilburn went on to serve the dean of the Faculty of
Science from 1970 to 1972 and pro vice chancellor between 1976 and 1979.
MU5
In 1966, Kilburn embarked on what was to be his last major computing project: the MU5. The Atlas had
been operational for four years, and the MU5’s main focus was to provide a computer architecture geared
toward the efficient running of programs written in high-level languages. The MU5 was conceived as a
range of three machines—a small inexpensive computer, a high-spec scientific computer with 20 times the
throughput of the Atlas, and a multiprocessor—but only the second was actually developed.
15
The MU5’s
original design proposal was set out in 1968 at the Edinburgh International Federation for Information
Processing (IFIP) conference in a paper authored jointly by Kilburn, Derrick Morris, Jeff Rohl, and Frank
Sumner.
16
An interesting technical aspect of the MU5 was the associative name store in which frequently used
scalar variables would automatically reside in a fast cache store. Morris explained, “This was as a result of
an analysis of the Atlas software, especially the instruction code. We learnt something about the frequency
of use of operands and control structures. The order code accommodated string functions and vector
functions.”
15
The university secured the cooperation of International Computers and Tabulators (ICT), which made
construction facilities available at cost and provided five staff to work on the project. The university’s
relationship with ICT and the potential it created to benefit the company persuaded the Science Research
Council (SRC) to assist the project by awarding the university a £630,000 grant over a five-year period.
It was a fruitful collaboration. However, an initial failure by International Computers Limited (ICL)
(which by then had merged with ICT) to acknowledge the extent to which the MU5 had influenced their
2900 series concerned the SRC, outraged Kilburn, and led to a long-running dispute that was not fully
settled until after Kilburn’s retirement in 1981.
Retirement
In order to spend more time with his wife, Kilburn retired early at age 60. Unfortunately, Irene Kilburn
died just two weeks before his planned retirement.
17
After that, he continued to spend one day each month
in his old department, but the majority of his time was spent with his son and daughter, gardening, playing
the piano, and following the Manchester United Football Club. He died in Manchester on 17 January 2001.
References and notes
1. M. Wilkes and H.J. Kahn, “Tom Kilburn CBE FREng. 11 August 1921 – 17 January 2001,” Biographical
Memoirs of the Fellows of the Royal Society, The Royal Soc. London, 2003, pp. 283–297.
2. Geoff Tootill was Christ’s College representative for the New Pythagoreans and president of the Archimedeans,
and Gordon Welchman was a student at Sydney Sussex College and honorary vice president of the New
Pythagoreans.
3. D.P. Anderson, “Max Newman: Forgotten Man of Early British Computing” Communications of the ACM, Vol.
56 No. 5, Pages 29-31
4. Kilburn and Tootill were also students in some of Newman’s classes.
5. G. Bowker and R. Giordano, “Interview with Tom Kilburn,” IEEE Annals of the History of Computing, vol. 15,
no. 5, 1993, pp. 17–32.
6. Originally trained as a chemist at Leicester and a physicist at Cambridge, Charles Percy Snow (Baron Snow of
Leicester) was mid-way through his four years of service as technical director of the Ministry of Labor at the time
of this talk.
7. T. Kilburn, “From Cathode Ray Tube to Ferranti Mark I Resurrection,” Resurrection: The Bulletin of the
Computer Conservation Society, vol. 1, no. 2, 1990, pp. 16–20.
8. D.P. Anderson, “Was the Manchester Baby Conceived at Bletchley Park?” Proc. Electronic Workshops
Computing, British Computer Soc., 2007; http://www.bcs.org/upload/pdf/ewic_tur04_paper3.pdf.
9. D.P. Anderson, “The Contribution of M.H.A. Newman and his Mathematicians to the Creation of the Manchester
‘Baby,’” The British Society for the History of Mathematics Bulletin, vol. 24, no. 1, 2009, pp. 27–39.
10. D.P. Anderson, “Patrick Blackett: Providing 'White Heat' to the British Computing Revolution” Communications
of the ACM, Vol. 56 No. 12, Pages 26-28
11. Actually, the process involved developing a series of successively more complex prototype machines.
12. E.M. Dunstan also joined in 1954 and worked on drum storage.
13. S.H. Lavington, A History of Manchester Computers, 2nd ed., British Computer Soc., 1998.
14. S.H. Lavington, “The Manchester Mark I and Atlas: A Historical Perspective,” Comm. ACM, vol. 21, no. 1,
1978, pp. 4–12.
15. D. Morris, “Early Computers at Manchester University,” Computer Resurrection, vol. 1, no. 4, 1992.
16. T. Kilburn et al., “A System Design Proposal,” IFIP Congress, vol. 2, 1968, pp. 806–811. See also R.N. Ibbett,
“The University of Manchester MU5 Project,” IEEE Annals of the History of Computing, vol. 21, no. 1, 1999,
pp. 24–33.
17. Sir Maurice Wilkes, interview by D.P. Anderson, Feb. 2009. An edited version of this conversation appeared
Communications of the ACM, Vol. 52 No. 9, Pages 39-42
Document Outline - Tom Kilburn: A tale of five computers
- Early Days
- War Service
- The Manchester Baby
- Mark I and Mercury
- Transistor Computers
- Muse and Atlas
- Department of Computer Science
- MU5
- Retirement
- References and notes
Dostları ilə paylaş: |