The upgrade to the UMIST DEC-10 is now complete with the arrival of the TU10 tape system. Interactive computing services are now available there and on the DEC-10 in Edinburgh. The PRIME 400 at Chilton should also be providing a service for users from 1 July 1977.
The SRC's subject committees will allocate computing resources to research workers in Allocation Units (AUs), arbitrary units which do not need to refer to any particular machine. In one year a total of 12,500 AUs should be available on the Edinburgh DEC-10, 10,000 on the UMIST machine and 4,000 on the PRIME at Chilton. Up to 15% of the available AUs can be allocated at the discretion of the Director of the Rutherford Laboratory. Much of this discretionary use will be for so-called "pump-priming". Groups who are applying for SRC grants will use it to test the suitability of the service for their projects while their applications are being considered.
Details of rationing and charging are given in section 4, below. More information is available from: Mr J E Hailstone, Atlas Computing Division, Rutherford Laboratory
The SRC is to pay for the enhancement of three mini-computers in universities at a cost not exceeding £50,000 each. The enhancements will consist of major improvements to the hardware so that the minis can offer a general multi-user interactive computing service. The computers are:
Also, the Department of Mechanical Engineering at Leeds University has been offered £60,000 by the University Grants Commission to upgrade its PDP11/45.
An order has been placed for the upgrade of the Nottingham machine to a PRIME 400 with 256Kbytes of core and 160Mbytes of disk space. The equipment is due to arrive at the beginning of June. It should be possible to provide an enhanced service fairly soon in view of the progress being made with the PRIME 400 at Chilton. Both PRIMES will have synchronous communication hardware and may be linked using PRIME's own communications software, PRIMENET. Orders have not yet been placed for the hardware for the remaining upgrades. The final configurations of the two PDP11/45s will be significantly different from each other.
The Rutherford Laboratory will contribute to the recurrent costs of running a service on all four machines and will control a share of the computing for the use of SRC grant-holders. One intention is to make the specialised software at each site available to whichever SRC-supported groups will benefit from it. For example, the extensive data-analysis software available on the PDP11/45 at Southampton should be of interest to research workers all over the country; the interactive graphics software on the PRIME at Nottingham - which supports an IMLAC PDS-4 refresh display - should get wider use outside the field of ergonomic modelling for which it was first provided.
Part of the acceptance tests of the GEC 4070 and the PRIME 400 was a requirement to re-run the interactive benchmark which had first been run as part of the tender. The benchmark consisted of a single script which was followed by 6 users simultaneously (5 on 1200 bits-per-second terminals, 1 on a teletype). The script consisted of six stages: two edits, one compilation and three program execution phases. Each user started at a different stage. Some improvements were expected as both manufacturers had supplied faster disk systems than those used in the earlier run of the benchmark; GEC had also made substantial changes to their operating system.
The GEC 4070 results showed an improvement of about 45% and the PRIME 400 about 20% in total elapsed times. Although the PRIME 400 processor is considerably more powerful than that of the GEC 4070, the elapsed times for the benchmark are now very similar. This is because both processors are sufficiently powerful, and the operating systems sufficiently efficient, to cause the benchmark performance on both machines to be limited by essential disk input/output activity.
The university groups which are collaborating with Rutherford Laboratory staff in assessing the machines have started work.
The accounting systems on the DEC-10s will measure the disk space, core occupation time, CPU time and terminal time used by each job. A weighted sum of these quantities measures the machine resources used in Computer Resource Units (CRUs) The number of CRUs will then be multiplied by a charging factor which reflects machine loading to arrive at the number of Allocation Units used. There are three levels of the charging factor (peak, standard and discount). When the machine is very lightly loaded at night and at weekends the discount level will apply. During normal office hours the peak level will apply, and on weekday evenings the standard level will apply. The peak, standard and discount levels are in the ratio 1.0:0.5:0.2.
The DEC-10s will use 4-week periods as the basis of their job accounting. Each user will be allocated a ration of AUs for each accounting period. Normally this will be done by averaging the number of AUs granted by SRC over the period of the grant. The number of AUs used by each job will be deducted from the ration. At the end of each accounting period any unused AUs will be carried over into the next period according to a formula which puts some constraints on the amount carried forward.
The nominal value of an AU is £15; operating installations can buy time at this rate. Government-funded research workers, if they are not supported by SRC or classed as pump-priming users, will pay the 'full cost' rate, which is £24 per AU; they will only be able to use the services at the discretion of the Director of the Rutherford Laboratory. Any use of the services by commercial bodies will be paid for at the 'market rate'. The allocation, charging and rationing procedures on the Chilton PRIME and the university minis will be based on similar principles.
The money values per AU will stay fixed until August 1977. They may need to be changed then but will not be altered more than once a year thereafter. The charging factor levels are unlikely to be changed; in any case they will not be changed more than once a year.
Two meetings were held on software for electric circuit design. Each meeting began with a description of the Interactive Computing Facility by Dr G Manning.
The meeting was held on 8 February and was chaired by Professor J V Oldfield (Swansea); 23 researchers attended. Talks were given on logical system design with fault test generation and on placement, routing and intercommunication software. Each talk was followed by a discussion period and the meeting recommended that a Special Interest Group should be set up.
The meeting was held on 10 February and was chaired by Mr G Freeman (CAD Centre, Cambridge); 27 researchers attended.
The first specialist talk demonstrated the need for both interactive and large-scale batch processing in computer-aided circuit design. The second covered simulation in electro-mechanical design. Each was followed by a discussion and the meeting recommended that a Special Interest Group should be set up.
These two meetings showed that researchers need increasing access both to fast interactive processing and also to the more powerful batch-orientated computers; this increase will continue. A long-term strategy is necessary to provide all that is required but much software exists and can be mounted for use in the short term. The Interactive Computing Facility is a step forward since it identifies areas of interest; assembles small expert groups to advise; and provides a substructure for better communication on all levels. The major requirements so far seem to be:
The GAELIC system for 'two-and-a-half dimensional' graphical editing of chip and mask layouts is available now on the DEC-10 at Edinburgh. The 'two-and-a-half dimensions' is convenient shorthand for a finite number of discrete planes. An early version of the Brunel University-Smiths Industries logic simulator HILO is also available now, on the 1906A at the Rutherford Laboratory.
As at the earlier meetings for users in other specialist fields, several participants emphasised the need for rapid dissemination of new software to avoid duplication of effort. Reports of the meetings will be prepared; anyone interested in getting a copy should write to the address given in section 1. The membership of the Special Interest Groups is listed in section 7.
The Interactive Computing Facility is controlled by SRC's Engineering Board and many of the users are concerned with engineering. However one of the first Science Board projects on the Interactive facility has begun an experimental service. The Crystal Structure Search Retrieval (CSSR) program is now being used on the Edinburgh DEC-10 by twelve chemistry and physics departments for a trial six-month period. CSSR is an online interactive program for retrieving information from the Cambridge Crystallographic Data Centre's base of crystal structure information on 16,000 organic and organo-metallic compounds. It provides fast retrieval of structural and bibliographic information and allows logical manipulation of retrieved material.
The user who wants structural information about a compound or group of compounds can define a 'fragment' or a 'ring' characteristic of part of the molecule he is studying. Each fragment consists of a single central atom and all its immediate neighbours; a ring is a fragment centred on a cyclic structure rather than a single atom. There is a special system for input of this information from a teletype. The computer replies by printing out the number of references in the data-base to molecules containing that fragment. If the number is large, the user can, repeat the process with another fragment and then get the computer to tell him which references are common to both fragments. He can keep on repeating until the number of references is down to manageable size or until he has specified the compound exactly; he can then inspect structural information in the entries to which references point. The information can be printed on a teletype or as a 3-D perspective plot on a graphics display unit.
The data-base is used in an 'inverted' form, which makes information retrieval a tree search and thus very fast. The references are associated into groups. The search process subdivides these groups until an exact match with the input 'fragment' or 'ring' is found. If the user wants information on a group of related compounds, he can deliberately stop short of defining a compound uniquely. This is useful if for example he wants to investigate a group of antibiotics with a useful property in common, in order to find one which does not have harmful side-effects associated with certain substituting groups.
The file inversion process is a very big job. It is run once every six months on the 360/195 at Chilton, to take account of new information. Shorter update runs are performed at three-month intervals.
The data-base also contains other useful data about the compounds, including some details of the methods used to determine their structures and full bibliographic information.
The SRC's Interactive Computing Facilities Committee (ICFC) has agreed to support the Special Interest Groups (SIGs), as recommended by the user's software meetings, for one year in the first instance. The following have been invited to be members of the SIGs; most have accepted but final confirmation is still awaited from a few:
The SIGs are to report on immediate software needs and to estimate the software effort required in the next five years. Eventually they will all include a representative of the Rutherford Laboratory; they are to receive technical and secretarial support from the Laboratory.
The ICFC will consider providing applications packages on the batch-processing mainframes if the SIGs make specific proposals. A working party, including a representative from each SIG, will define requirements for pre- and post- processor software on the interactive machines to access specific programs on the mainframes. The SIG for electromagnetic systems has already met.
In March a post at UMIST was advertised for a user-support programmer who would spend a large part of his time dealing with users at remote sites. The post was not filled and has recently been advertised again.
The post of Head of the Interactive Computing Facility has recently been advertised in the press. The appointment is at Chilton and is for a period of three to five years in the grade of Senior Principal Scientific Officer.
Four posts at Chilton for Research Associates have recently been advertised. These are three-year fixed term appointments; the successful applicants will work on applications software in close collaboration with the SIGs. Also, some permanent posts at Chilton are being advertised at the moment for applications programmers. They are in the grade of Scientific Officer or Higher Scientific Officer.
F10 FORTRAN and LINK overlay will be available very shortly at UMIST. It is proposed that the GINO graphics package and the NAG library of subroutines for numerical analysis should also be provided there. GINO-F has recently been made available on the PRIME at Chilton.
There is to be standardisation on terminals as far as possible. The standard VDUs will be Newbury Laboratories 7002s and the standard portable hardcopy units will be Transdata 705s. The standard graphics storage tubes will be Tektronix 4010s; work is still going on to select a standard fast serial printer. 50 terminals were ordered in 1976-7 and most are being sent out to universities. Proposals have been made to purchase a further 66 terminals; some are specifically requested to cope with the demand arising from the upgrades of university minis. The terminals are part of a "pool" of equipment managed by the Rutherford Laboratory and can be sent to different places when the need arises.
A small network-node computer, from the DAS-80 range produced by DEC, will be installed at Chilton. It will be used to 'forward link' terminal users in the south of England, including those connected via concentrators. to UMIST and Edinburgh. Two GEC 2050 minis may be installed in universities to act as concentrators
The UMIST and Edinburgh DEC-10s have been linked via Chilton using DECNET. A direct link is to be installed later in the year.
Dr P J Hayes of the University of Essex writes:
"In reply to Dr Andrew's comments on LISP and ALGOL-68 in issue No 2 of Rapid Response. I think he has the wrong notion of just how intricate a modern LISP system is. It is, indeed, fairly straightforward to write a basic LISP 1.5 interpreter in ALGOL-68. or indeed in almost any reasonable system programming language. But modern LISP systems such as INTERLISP or MACLISP are very large. sophisticated interactive programming environments. To engineer such a system from scratch would be a major undertaking. The AI programs we want to use will not run on a basic LISP 1.5 system."
"The time penalty involved in a purely interpretive system is, in any case, fairly prohibitive, typically two orders of magnitude compared to a good compiler."
"It has been suggested that one should use ALGOL-68 instead of LISP in interactive developments of large AI programs. Some groups have used ALGOL-68, I understand, but it does not seem to be a really useful thing because its control structures are so archaic as to make complicated co-routine programming virtually impossible."
The next issue will be published in September. In it I hope to include:
From time to time I will publish articles about particular projects using the Interactive Computing Facility. Contributions and letters from readers are always welcome.
