The Transition from Centralized to Distributed Computing. 1980-1992

Professor Harry Whitfield

Director of the Computing Laboratory
Head of the Computing Science Department
1980-1992


Organization of The Computing Laboratory

When I became Director of the Computing Laboratory on 1st January 1980, the Laboratory had about 120 staff. There were about 30 teaching and research staff, 40 academic related computing service staff and about 50 support staff.

In the period between Professor Page leaving and my arrival, the University had taken the step of separating the funding and reporting routes for the "teaching" and "service" sides of the Laboratory. The teaching was under the Faculty of Science and the Service was under the Computing Laboratory Committee.

Almost all of the graduate staff had functions which lay quite clearly in one or other of the two sides, the main exception being the Director himself. A few of the support staff had functions which covered the whole Laboratory.

The main advantage of housing these two activities in one Laboratory, with one Director, was that the Director could (when necessary) exert pressure on both sides to communicate and cooperate. In the eighties there was much to be gained from this approach, in the rapid transfer of technology from the research to the service domain. I will cite examples of this later in this talk.

Ten years later, the Laboratory still had over 120 staff, but the distribution was quite different. There were now about 45 teachers, researchers and computing officers supporting computing science activities, but just over 30 graduate staff in the computing service. The number of other support staff was still about 50, but the great majority of these were supporting computing service activities.

During the same period, the number of students in computing science had increased five-fold and the number of computers in the University had grown by about three orders of magnitude.

Towards the end of the decade the Crombie Committee proposed the formal division of the Computing Laboratory into a Department of Computing Science and a University Computing Service. Although most of the Laboratory staff were opposed to this division, and Senate initially rejected the proposal, the division eventually took place on 1st August 1992.

Staff of the Department and the UCS still share a common room and cooperate on a number of technical and domestic issues, but the drift apart which we predicted is clearly to be seen, despite the excellent contacts between staff who predate the division. Computing Science Staff are now under so much pressure from increased student numbers and the need to do well in research assessment that they have little time (and no obligation) to help their service colleagues. On the other hand, the Computing Service is so constrained by market developments and lack of staff time, that any form of independent development seems neither worthwhile nor possible. We are truly in the age of "Commodity Computing"

NUMAC and Computer Board Replacement Cycles

As Director of the Computing Laboratory at Newcastle, I was also Director of NUMAC, the organization set up by the Universities of Durham and Newcastle to provide a joint computing service.

During my time as Director, life seemed to be dominated by Computer Board replacement cycles, which were originally at ten year intervals and then at seven year intervals with a mid-term review.

In 1980, it became clear that NUMAC was no longer to be a National Resource Centre, and we could get on with providing the best service we could for Durham and Newcastle Universities.

Our old IBM 360/67 machine was switched off in July 1980, and the IBM 370/168 continued as our main Time-Sharing service with MTS (The Michigan Terminal System). At that time our network supported 200 terminals and microcomputers. We had experimental external networking with EPSS, later to become PSS. Planning for the 1980s was started, which resulted in the report "Computing at the Universities of Durham and Newcastle in the 1980's".

In 1982 an IBM 4341 was installed at Durham. This machine was capable of supporting 40 terminals and NUNET gave access to the two MTS services and 6 PDP11 based services (with two machines in Durham). We had a connection to PSS and we were planning a graphics service based on a Vax 11/780.

In 1984 we were connected to JANET, but only for batch services.

In 1985 the Computer Board provided £1.8M for NUMAC and we installed an Amdahl 5860 (with 40Mb of main memory) at Newcastle and an Amdahl 470/V8 (with 16 Mb of main memory) at Durham. We continued to run MTS. Three years later, the Durham Amdahl 470/V8 was replaced by a second Amdahl 5860.

We introduced the New User Interface to Terminals based on the Yellow Book Transport Service (YBTS) and this gave direct terminal access to JANET. At Newcastle we set up the first cluster room with 40 PCs.

In 1989 we began planning for the demise of MTS. NUNET, too, had to be phased out by 1992 as we needed MTS to support it. We installed our first main service machine to run UNIX (Gould NP/1).

Our next procurement would be for a large number of workstations supported by a central server, but I will leave that story to Dr. Russell.

Screen Editing

One of the things which concerned me most when I came to Newcastle was the lack of penetration of the network and terminals into departments. I was also greatly concerned that only directly connected terminals had any form of screen editing, and that our strategy for network expansion seemed to preclude it.

Discussions between Alan Hunter, Dennis Russell and myself led to the what we now call the "Cursor Box Algorithm". In turn, this enabled us to provide the SC Screen Editor (implemented by Alan Hunter and Nigel Hall). The work eventually led on to the Simple Screen Management Protocol (SSMP), which was blessed by the Computer Board and became the Fawn Book Standard, and the design and production of the Fawn boxes which were able to support the Curlew Screen Editor and the EMU mailer.

I would like to tell you more about the Cursor Box Algorithm, because this was a case where control of our own networking software gave us an enormous return for little effort.

Cursor Box Algorithm

The Computer Board and Network Protocols

By contrast, most of the decade was bedevilled by the imposition by Government (via the Computer Board) of the wrong network standards. Both the UK Government and the European Commission were pushing ISO standards (in the hope of giving UK and European industry an advantage), whilst our academic partners in the USA were developing the TCP/IP family of protocols. The Computer Board used its procurement rules to force companies wishing to sell to UK universities to implement the JNT Coloured Book Standards which fitted the ISO approach.

We, in NUMAC, were obliged to bring our network into line with the Computer Board's recommended protocols and did, indeed, help to develop and implement many of the protocols for the UK community. We had little choice but to conform, but, although initially there were some marginal gains, it soon became clear that the returns were minimal compared with that being achieved by the use of TCP/IP in the USA. The Board held to its policy long after it had ceased to be viable, causing a massive waste of time and effort in the UK networking community.

Towards the end of the decade, it was quite apparent that UK universities would be cut off from important developments in networking if the Computer Board was not persuaded to change its policies. By then, I was a member of the Computer Board and became Chairman of the Network Advisory Committee. I was able to push the NAC, the Computer Board and the JNT into allowing a multi-protocol network, where TCP/IP protocols would be carried on top of the ISO protocols. Within a few months, use of IP was dominating the network and the Computer Board accepted the inevitable. Once the policy hurdle had been overcome, Dennis Russell played a major role in helping to develop IP networking in the JANET network.

Interplay between Research and Service Developments

In 1982, as part of our Distributed Systems research, we were developing the Newcastle Connection. This was a software system which implemented (on a set of machines running the Unix operating system) a Coherent Distributed System which enabled users to access files on any of the machines and run programs without being directly concerned about their physical location. This work was going on at the same time as we were planning our computing strategy for the 1980's and greatly influenced our thinking.

Although we did not dare, in the end, to move to fully distributed computing in the 1985 procurement, the seeds were already sown.

In 1985/86 Computing Science won a $1M gift of Xerox Workstations and Servers. This led immediately to the installation of an Ethernet in the Computing Laboratory, but more importantly gave many of us direct experience of working in a sophisticated distributed client-server environment. We had earlier had a gift of BLIT terminals from Bell Laboratories, which gave a few people experience of bit-mapped displays.

In 1987/88, after another standards battle had been resolved, the computing service installed 13 km of fibre for the Ethernet which forms the backbone of the present Campus Network.

In 1987, the MTS community set up the Workstation Integrated Network Environment Project (WINE or Intersystem Project), which I directed for about three years. Although we decided there was no point in trying to produce software, it did give us a lot of insight into the problems of heterogeneous environments.

In 1986/87 as part of our Reliable Distributed Computing research, work began on the ARJUNA system. This system has been used (in anger!) at the start of each of the past four academic years to run the University Registration System.

Similarly, Professor Lee's work on Parallel Processing using Shared Memory Multiprocessors led (in 1988) to a £0.5M grant from the DTI and the installation of a powerful machine from which the whole Newcastle community has been able to benefit.

Teaching

Computing Science Teaching has been dominated by change, both in the subject matter we teach and in the types of courses and students.

Our Computing Science courses were restructured in 1983, in 1989, and more recently in 1996. In 1980, with four other departments, we set up a new degree in Microelectronics and Software Engineering. Our major partner in this was (and still is) the Department of Electrical and Electronic Engineering, with whom we also have joint research projects. In 1984, we started the 4 year MEng Degree in Microelectronics and Software Engineering.

The number of students has grown enormously. In 1981, we awarded 13 Single Honours Degrees in Computing Science. In 1985 we awarded 19, in 1987 there were 32, and in 1988 there were 41. In 1988 there were 20 MEng graduates, too.

In 1997, we admitted well over 100 students to our Single Honours course. Our teaching staff now numbers 26, and our FTE student numbers are nearing 400. We have had a five-fold increase in student numbers with a 50% increase in staff.


Contents Page - 40 years of Computing at Newcastle


The Transition from Centralized to Distributed Computing, 20 October 1997