Jump Over Left Menu
Annual Report 1970
Computer Systems and Applications
THE NEW CENTRAL COMPUTER - IBM 360/195
Approval was received on 10th November 1970 for the purchase of an IBM 360/195 computer, with peripheral equipment, at a total cost of Â£3.6M. This sum includes the cost of the building extension estimated at Â£122K.
An Advisory Committee has been formed to advise the Director of the Rutherford Laboratory on Computer Policy. Membership of the Committee represents the respective interests of the Rutherford, Daresbury and Atlas Laboratories. The computer will be available not only to the Rutherford Laboratory but will also be used for off-loading work from the Daresbury Laboratory and computers used at universities for bubble chamber film analysis. Computing facilities for nuclear structure groups and Elementary Particle Theorists will be authorized through the Rutherford or Daresbury Laboratories. A small proportion of time (10 to 20%) will be authorized through the Atlas Laboratory for its own users.
At its first meeting on 18th January 1971, the Advisory Committee approved the following configuration for the computer:
- IBM 360/195 CPU + 2 Mega-byte Store
- 4 Line Printers
- Card Reader/Punch
- Block Multiplexor
- Fixed Head File
- 2314 Disk Store
- 3330 Disk Store
- 8 Ã— 1600 bpi Magnetic Tapes
- 2 Ã— Dual Density Magnetic Tapes
- 2 Ã— 7-Track Magnetic Tapes
All of the peripherals (except the Line Printers and Card Reader) are of IBM 370-type.
The Central Computer (CPU, Store, Multiplexors) is scheduled for delivery on 1st November, 1971. This part will be installed and tested without interruption of the 360/75 operation during November and early December. When commissioning is complete the existing 360/75 peripherals will be switched over. The down-time is estimated to be 5 days during the latter part of December.
The 370-type peripherals will not be available until February 1972. A phased programme of installation has been prepared causing as little inconvenience to users as possible during the period February to April 1972.
Work has started on the new southern extension to the computer wing. The new computer area should be available for occupation by mid-August, 1971.
THE IBM 360/75 CENTRAL COMPUTER
The small 2311 disks have been replaced by two 8-drive 2314 disk sets. This change has eased maintenance and development, and given better throughput and better facilities for user data. A second console (1052 plus 2150) has been added, primarily to facilitate tape and disk mounting. Additional parallel data adaptors have been put in the two 2701 data control units, for attachment of the terminal IBM 1130 and DDP 516 computers as extra satellites. An IBM 2702 Multiplexor to be commissioned shortly will provide another 12 access points to the 360/75. Six of these will go through the Rowstock Telephone Exchange. The IBM 1130 has been delivered and accepted, and communication between the 360/75 central computer and all its satellites is now reliable. Two visual display units (Tektronix storage tubes) have been attached to the DDP 224 satellite. The data-link to the London Institute of Computer Science has been made bi-synchronous.
Four-shift working is now normal and 4,000 jobs per week were being done by Operations the end of the year. During the last quarter the overall machine efficiency was about 98%, with CPU utilisation of 88%; 9-track magnetic tapes were mounted 33,000 times. Performance during 1970 is shown in figures 128(a) and (b).
Figure 128(a). Total number of Jobs on Central Computer during 1970.
Figure 128(b). Central Computer machine statistics during 1970.
Release 18.6 of the IBM Operating System was installed in September. After some initial troubles and insertion of PTF's (Program Temporary Fixes), it was decided to obtain in advance some modules of release 19 and retain some of 17, and the resulting system is very reliable. The main reason for introducing 18.6 was software support for the second control console. The sub-system HASP-2 was introduced in December, mainly because of its facility for remote job submission from the interactive terminals controlled by the ELECTRIC program.
Studies on changing the operating system from multi-programming with a fixed number of tasks (MFT) to multi-programming with a variable number of tasks (MVT) are well advanced, and methods of avoiding memory fragmentation (aggravated by very long jobs such as on-line automatic measuring machines) have been developed.
The CERN and AERE libraries of programs and subprograms have been made available, and facilities developed whereby users can incorporate items from these libraries in their programs. Hitherto subprograms of general use written at the Rutherford Laboratory had been incorporated in the Fortran automatic-call library and the IBM Scientific Subroutine Package. These latter have now been left as they are issued and the Rutherford-written routines form a new library on their own.
The Multi-Satellite Control Program MAST-DAEDALUS
The MAST program, for controlling several satellite computers by the DAEDALUS method, was written during the year and installed in December. It now runs whenever the computer is on, at present handling 3 satellites (DDP 224, IBM 1130, DDP 516. The control program that runs in the DDP 224 satellite is now stored in relocatable form in the 360 disk store, whence it is fetched by a small loader as required. The programs which control the various devices attached to the DDP 224 are also stored in this way, and the space occupied by them in the DDP 224 is freed when the user goes off-line. Data buffer-space is also acquired and released dynamically. Messages destined for fast or slow devices are now handled differently and output data can be buffered if requested.
The present version of MAST is designed to be secure against misuse, and to give the various concurrent users complete protection from each other. DAEDALUS code in the DDP 224 and IBM 1130 is virtually complete for the present level of design, and is being written for the DDP 516.
On-Line Computer Applications
AUTOMATIC FILM MEASURING
During 1970 HPD I measured 210,000 events on film from four experiments undertaken by the Bubble Chamber Research Group. The film was exposed at CERN 2 metre chamber (some as long ago as 1966, but originally intended for a different experiment).
Beam Design Programs
An interactive visual-display (graphics) program is being developed to complement the existing and widely used beam optics programs, TRAMP and IPSO FACTO. Initial definition of the beam is based on TRAMP, but thereafter the flow of the program has been completely changed to simplify the task of adding extra facilities as their need becomes apparent.
Output is routed via the DDP224 satellite computer either to the Computek display for graphical output, or to an adjacent typewriter. Communication with the program is via the typewriter keyboard. Modifications to the beam-line or automatic numerical matching can be performed and the resultant effect on stored particle trajectories can be displayed rapidly. This interactive aspect of the program is its key feature and should greatly facilitate the task of beam design, and the development of tuning. procedures for existing beam-lines.
EXTENSION OF ON-LINE SYSTEMS
New Satellite Computers
In addition to the DDP 224, through which all on-line users have communicated with the central computer so far, two new satellites will soon be in full operation. They are a DDP 516 dedicated to HPD II, and an IBM 1130, which will be situated in the Nimrod Experimental Area, and function as a remote terminal. In due course the IBM 1130 used for gathering data from the scanning and pre-digitising machines will also be linked to the central computer.
A new multiplexor is being built for the DDP 224, to allow simultaneous operation of the existing typewriter consoles and connection of up to a total of 16 similar devices.
The terminal IBM 1130 is at present connected to the 360/75 by a short distance link, and the magnetic tape unit, line printer and card reader are being commissioned using an operating system similar to that on the DDP 224. A unit has been added to the terminal computer to allow connection of a further eight devices. One is a multiplexor, connecting users in the Experimental Area via their CAMAC systems. Another is an 880 kilo-bit per second serial link, capable of operating over distances exceeding half a mile, which will be used when the terminal computer reaches its site in the Experimental Area (500 yards from the central computer).
Once established there, the terminal will allow users to submit jobs or data (cards or 7 -track magnetic tape) to the 360/75 and receive back output on an 80 lines per minute printer. It will also be possible to connect various user terminals (teletypewriters, small computers, or visual displays).
A direct link between the DDP 516 computer (controlling HPD II) and the central computer has been established and is operating satisfactorily. Current hardware, including a newly-built multiplexor for the direct control feature of the IBM 360/75, allows the connection of up to six satellite computers.
Bill Walkinshaw, Head of C and A, February 1970
The ELECTRIC Program
A program named ELECTRIC has been developed to provide remote job entry to the central computer from typewriter consoles which will be distributed around the Laboratory. It operates under the DAEDALUS message switching system and gives a general service. A user may create and edit files containing the text of programs, job control statements or data, and may submit a file as a job for batch execution. The program also allows monitoring of the subsequent progress of the job.
In addition to these normal facilities for conversational remote job submission, the program allows the user to hold files in the form of text with associated editing instructions. The files can then be used in their original text form or as updated by their associated edit instructions (or sub-sets of these instructions). A comprehensive set of manipulations of both text and edit instructions is available.
ELECTRIC is also intended to help in the following problem situations which often arise in computing:
- Keeping several slightly different versions of the same big basic program.
- Supplying new versions of programs to outside users who have themselves modified the previous version.
- Difficulties when more than one person is working on a program.
The system has been in use with a limited number of typewriter consoles since mid-1970.
DATA ANALYSIS SOFTWARE
Bubble Chamber Programs
Maintenance and modifications of the existing Road Guidance system, and development of the Minimum Guidance system; have been the main activities this year.
Automatic Track Matching was introduced early in 1970, and has been used in processing all of the three experiments measured since. No other changes of principle have been made to the production version of the date processing program, though there have been numerous small changes.
Data safety has been considerably improved by an automatic file protection program now in regular use. Tapes containing data accumulate at several stages of HPD production measurement, e.g. digitisings prior to week-end HAZE runs, and geometry input prior to merging with output from the patch-up system, which may not be available for a week or two. To prevent tapes being accidentally over-written a record is held (on disk) of the current use of each HPD tape, and programs which output data to tape have been modified to check that the tape is available for use, and to update the record. A list of free tapes can be generated at any time.
In the Minimum Guidance system, two major problems are the number of digitisings produced (two to six times as many as in Road Guidance) and the CPU time taken by the filter program (about three times Road Guidance). A method under test would reduce the digitisings by defining, at the measuring stage, a cone containing all tracks of an event, and gating-out other digitisings.
A sample of events from the low-energy K- P experiment is being used for Minimum Guidance tests. The main problems arising in matching tracks found in the three views have been due to non-interacting beam tracks close to the production vertex, and (in multi-vertex events) tracks being picked up by the filter program at more than one vertex. For example, the track from a L decay may well be picked up at the production vertex and at the decay vertex.
A patch-up system using the IDI display and light pen has been developed, and has been successfully applied to rescue more than half of the failed events in a test batch. The region of the picture close to the vertex is displayed, but without digitisings of tracks already successfully matched. The operator marks three points on tracks chosen (as in Road Guidance), and filtering then proceeds as in Road Guidance.