The SERC Council, at its meeting on 21 January 1982, agreed to the purchase of a 16 Mbyte, 16 channel IBM 3081D to be installed at Rutherford Appleton Laboratory (see January 1982 issue of FORUM). Delivery will be in June. Capital funds, allocated to Central Computing in future years, have been brought forward to reduce the very high maintenance and energy costs of the 360/195s.
It is proposed that one 360/195 is removed when the 3081 is delivered and the other one removed a few months later. (The vast amount of space freed in this way will be further supplemented by the removal of the Univac 1108 which RAL currently runs on behalf of NERC).
After acceptance, the 3081 will simply replace one 360/195. It is planned after a month to swop the 3081 and the 3032 front-end. The 3081 will then run an enlarged front-end CMS service and the 'FEM' (the front-end MVT system) will also be able to run much more batch work than at present. It is because the front-end facilities are subject to the greatest pressure that this tactic is being adopted.
In September, by which time the 3081 should be tuned to its new role, the rear-end 360/195 should be removed. There will then be a net loss of bsteh capacity of about 1500 - 2000 hours per year. However, some capital funds remain allocated to Central Computing in financial years 83/84 and 84/85 and a continuation of the replacement plan is expected.
None of these manoeuvres should have any effect on the user's view of the system other than a change in service levels. The 3081 is a modern machine and gives the opportunity to move on to a supported operating system, thus putting the RAL Central Computing on a very stable basis.
The following is a brief description of the way in which AUs are calculated and how they are used in a normal terminal session. The algorithm currently in use for calculating AUs from the resources used is:
AUs = K C(t) [A × cpu secs + B × connect secs + C × spooled i/o + D × non-spooled i/o]
where:
A = 1.00000, B = 0.00060, C = 0.00035, D = 0.00800, K = 0.00410, and C(t) is given in the following table:
TIME | C(t) | TIME | C(t) |
---|---|---|---|
00.00 - 08.00 | 0.1 | 13.00-18.00 | 1.0 |
08.00 - 09.00 | 0.4 | 18.00 - 22.00 | 0.4 |
09.00 - 12.00 | 0.8 | 22.00 - 24.00 | 0.1 |
12.00 - 13.00 | 0.6 | w/e public holidays | 0.1 |
The charge rate for jobs submitted to CMSBATCH is as in the table, except that it never rises above 0.4.
The number of AUs used at any time in a terminal session can be determined by issuing the AUS command, which also returns the CPU time, connect time, spooled i/os and non-spooled i/os used for the calculation. The current charge factor and ration for the account can be obtained using AUS CF and AUS RATION. As can be seen, the most expensive time is on weekday afternoons. Issuing the AUS command periodically can help users to find out where they, in particular, are using most AUs .
The most significant contributors to the AU charge for a session are normally the CPU time used and the non-spooled i/o. Non-spooled i/o is any transfer of data which does not involve the spool, for instance reading or writing files on disk. Spooled i/o refers to transfer of data between devices which do use the spool, ie virtual and real printers, punches and readers.
The most common commands which use AUs heavily tend to be large compilations, assemblies and running jobs. Optimisation in compilations can increase the cost dramatically and it may well be worth requesting no optimisation while debugging. The Fortran GI compiler is much cheaper than the HX compiler when the HX compiler is requested using full optimisation (which is the default). In a test using a program of roughly 450 lines of Fortran, the FORTHX compiler was more than four times as expensive as FORTGI. At present there is no optimisation available for the FORTVS compiler and it compares roughly with FORTGI.
Large compilations and jobs are best run in the CMSBATCH machine which has a lower charge rate than normal CMS machines during the day. The output is returned via the spool and can be examined using the BROWSE command. This is much cheaper than reading the returned files on to a disk to look at them. Any files which require printing can be transferred directly to the real printer and then the only non-spooled i/o involved in processing those files is that done by the BROWSE command.
Another expensive command is LOAD for large TEXT files. When possible, a MODULE file should be generated using LOAD and GENMCD so that the loading is only done once and subsequent runs can use the module.
If a high proportion of large files exist on a disk, it may be worthwhile to re-format the disk to use a larger block size. This cuts down the number of AUs used in reading the files since a smaller number of i/os will be required. However, the number of blocks available on the disk will be reduced so it is not useful for a disk of small files.
Most commonly used CP and CMS commands are comparatively inexpensive in AUs, for instance QUERY and SET commands, MSG, LINK and ACCESS. The AU cost of LISTFILE, BROWSE, HELP and commands such as COPYFILE, COMPARE, TYPE and PRINT obviously depend also on the size of the file being processed. A typical XEDIT session uses AUs in rough proportion to the number of XEDIT subcommands issued. A very rough guide is 0.1 to 0.5 AUs for a half hour of XEDIT at the maximum charge rate.
A good idea of where AUs are being used in a particular machine can be obtained using CP SPOOL CONSOLE START * to spool the console to the users own virtual reader. The command CP SPOOL CONSOLE STOP CLOSE should be used to close the file. If the AUS command is issued periodically throughout a session, the console log can be examined later to find out which were the most expensive commands.
Two major programmes of work are scheduled for the next 6 months. These are conversion from HASP to VNET and in addition a programme of workstation upgrades.
The initial 'home1 installations of VNET are just about complete. As expected a small number of implementation problems have showed but they have been identified and fixed. RAL workstations will now be converted to VNET at a rate of approximately one per week, with a site specific User's Manual being produced for each change. Any comments on VNET (whether problems or usage) should be sent to Program Advisory Office - CMS ID=US, ELECTRIC ID=US
In the ICF area a small number of sites are being given larger machines and the redundant machines will be re-configured to provide a larger and improved workstation facility at certain GEC2050 sites. The displaced GEC2050 hardware will be used to enlarge non-networked GEC2050s which currently cannot connect to the SERC network. At other sites existing ICF facilities will be modified where necessary to accommodate an existing population of GEC2050 users and the GEC2050 removed. The following table indicates those sites for which changes have been agreed:
Site | RAL connection | Current equipment | Upgrade |
---|---|---|---|
Bangor | SERCNET | GEC2050 | DEC1O-Gateway |
Southampton | SERCNET | GEC2050 | GEC4070 |
Reading | SERCNET | GEC2050 | GEC4070 |
Durham | SERCNET | GEC2050 | GEC4070 |
Bristol | SERCNET | GEC2050 | ICF-GEC4090 |
Sussex | SERCNET | GEC2050 | ICF-PRIME |
Surrey | HASP-M/L | GEC2050 | ICF-PRIME |
Westfield | HASP-M/L | GEC2050 | SERCNET-GEC2050 |
Edinburgh- Univ-Phys | HASP-M/L | PDP11/40 | SERCNET-GEC2050 |
Exeter | HASP-M/L | GEC2050 | PDP11/40 |
It has been decided to extend the SERC Network with Packet Switch Exchanges to be installed in Edinburgh, London, Cambridge and CERN. The main purpose of these PSEs is to rationalise complex operational circuits where a number of otherwise independent connections share the same physical line. Ideally each PSE will be independently connected to at least two other PSEs in the Network. A further potential effect is to produce considerable long-term savings in Private Wire rental costs. This will in some way affect most users, in that their connection point to the Network will be changed. Service implementations of terminal protocols will need to incorporate the new Network addresses. Schedules for the changes will be circulated as the details become clear. Network access to CMS and VNET is now handled by a virtual machine called VMNCP rather than DKNCP running under FEM and handling HASP and ELECTRIC traffic.
New PACX software is in use which enables a message stored in PACX to be displayed on making a connection. In addition a number of ports have been changed from ELECTRIC access to CMS and made up to 4800 bps. This process will continue as the use of ELECTRIC goes down.
It is intended to standardise on 4800 bps line speeds in CMS and ELECTRIC. However there are technical modifications that need to be made to the communication lines. A few slots will be retained at other speeds for those terminals which require them.
When the IBM 3081D begins service it will provide part of the MVT batch service. The restriction of 1.4 Mbytes maximum REGION size could in principle be eased. Likewise there will be less need to penalise jobs with large core requirements. This will be of particular benefit to programs with large Overlay Structures. Users with programs of this nature are requested to contact J C Gordon in User Interface Group (CMS ID is JCG, ELECTRIC ID is PY). See also the item on Trial MVS.
As part of the move to the MVS operating system on the IBM 3081D (see January issue of FORUM, item 2), we will be introducing a trial MVS system, hopefully in the latter part of this year. In order to identify users who would benefit most from time on this system, we wish to contact groups whose work consists of large core jobs but with limited input/output. In particular we would like to hear from anyone who finds the size (1.4M) of the 195 a restriction. Potential users should contact J C Gordon at RAL, exts 6574 and 6111. CMS ID=JCG or ELECTRIC ID=PY.
Too many files are being left on the Spool for long periods. Users are urged to remove them quickly. Unclaimed files are erased if not removed in reasonable time.
The following items of interest have been taken from the Central Computing Site Users Meeting held on Wednesday 6 January 1982.
List of articles in FORUM 16
16.1 Introduction 16.2 Changes at RAL 16.3 Central Computer Procurement 16.4 IBM VS FORTRAN Program 16.5 Extract from minutes of PRIME User Meeting 16.6 Questions raised at CCR meeting, 13/7/81 16.7 Courses 16.8 Diary 16.9 Any offers - information needed about tapes 16.10 Computer Statistics 16.11 Index
The following dates have been decided on for DECsystem-10 Users Committee meetings during 1982:
The time and place for these four meetings is 10.30 am at the James Clark Maxwell Building, King's Buildings, Edinburgh.
There are four levels of support, as follows:
The packages listed below are those afforded some level of support at the end of December 1981. Initial queries may be directed to the Program Adviser, who will either deal with the query, or where appropriate, put the user in touch with the appropriate expert. No commitment is implied on anyone to continue support of a particular package but certain packages are offered with minimum notices of withdrawal periods specified. If users have any suggestions as to additions, deletions, corrections or changes which should be made to this list they should contact D. H. Trew, User Interface Group, Computing Division.
The following abbreviations are used in the list below:
PACKAGE | LEVEL | SUPPORT | DESCRIPTION |
---|---|---|---|
ALCHEMY | Min | DL | Quantum Chemistry |
ALGOL | Min | CD | Compiler |
APPLE | Min | BCRG | Direct Channel Partial Wave Analysis |
ASAS | High | SNS | Finite Element |
ASCOLI | Min | BCRG | Three Body Partial Wave Analysis |
ASSEMBLER (F) | Max | CD | 360 Assembler |
ASSEMBLER (VS) | High | CD | 360 Assembler |
ASTAP | Min | Tech | Statistical Analysis Program |
ATMOL | Min | DL | Quantum Chemistry |
BABBAGE XREF | Std | SNS | Formatted listing of BABBAGE programs |
BCPL | Std | CD | Compiler |
BDMS | Min | Durham | Berkeley Data Base Management System |
BERSAFE PH2 | High | TECH | Finite Element |
BERSAFE PH3 | High | TECH | Finite Element |
BERDYNE | High | TECH | Finite Element |
BMDX72 | Min | CD | Bio-Medical Statistics Package |
BUEDDY | Std | Tech | 2-D Eddy Current Program |
CAMAL | Min | CD | Symbolic Algebra |
CAPSTAN | Min | AERE | Critical Path Analysis |
CASORT | Std | CD | Sort/Merge |
CCOPY | Std | CD | Lists/Copies Sequential Files |
CLIST | Std | CD | Lists or Copies Source Files |
CERN Library | Std | CD | Subroutine Library |
CFD/CFDX | Min | CD | Translator generates ASK for ILLIAC IV |
COBOL(E,F,ANS) | Std | IBM | Compilers |
COCOA | Min | CD | Text Processing |
COPYTP | Std | SNS | Copies GEC 4080 tapes |
CPC Library | Std | CD/CPC | Subroutine Library |
CSL | Min | ICF | Simulation |
CSMP 1 | Min | CD | Continuous System Modelling |
CSMP 3 | Min | CD | Continuous System Modelling |
DIRMAINT | Std | CD | VM Directory Maintenance |
DMS | Std | CD | Display Management System |
DSKSOL | Std | Tech | Solves large sets of linear equations |
ELARKIVE | High | CD | Electric File Archive Facility |
ELMUG/ELFR80 | High | CD | Output ELECTRIC Graphics Files |
ELOUTPUT | High | CD | ELECTRIC Files to Printer/Punch/Tape |
ELDIRE | High | CD | Batch Access to ELECTRIC Directories |
ELSEND | High | CD | Batch Access to ELECTRIC Files |
ELUSER | High | CD | Batch Access to ELECTRIC |
ELECTRIC | Max | CD | Text Editing |
FAMULUS | Std | CD | Text Sorting and Indexing |
FAPLOT (ENPLOT) | Std | CD/HEP | Histogramming Package |
FELIB | High | TECH | Finite Element Library |
FEMGEN | High | Tech | Finite Element |
FEMVIEW | High | Tech | Finite Element Routines |
FORTRAN G1 | Max | CD | Compiler |
FORTRAN H Extended Plus | Max | CD | Compiler |
FORTRAN VS | Std | CD | Fortran 77 Compiler CMS only |
FOWL | Std | CD | Monte-Carlo Phase Space Program |
GENSTAT | Min | CD | Statistics package |
GEOMETRY,KINEMATICS & ORACLE |
Min | BCRG | Film Analysis Programs |
GEROFF | Min | CD | Text Processing System |
G-EXEC | Max | NERC | Relational Data Base and General Data Handling System |
GFUN | Std | Tech | Magnet Design |
GINO-F | Std | CD | Graphics |
GPSS | Min | CD | Simulation |
GRAPHICS/SUMX | Min | BCRG | SUMX with Graphics Options(MUGWUMP) |
Harwell Library | Std | CD | Subroutine Library |
HBOOK | Std | HEP | Histogramming Package |
HPLOT | Min | HEP | Graphics (part of HBOOK) |
HYDRA | Std | HEP | Program Management with Dynamic Memory |
IBM Utilities | Std | CD | (some) |
ICCG | Std | Tech | Solution of Sparse linear equations |
INFOL | Std | HEP | Database Report Generator |
JSPLOT | Min | BCRG | Histograms/Scatter Plots via MUGWUMP |
KWIC360 | Min | CD | Text Processing Index System |
MAST | High | CD | Message Switching |
MINUIT | Std | Oxford | Minimizing Package from CERN |
MORTRAN | Std | HEP | Structured FORTRAN pre-processor |
MTUT | Std> | SNS | Initialises and checks GEC 4080 tapes |
MUGWUMP | High | CD | Graphics Package and Filestore |
NAG Library | Std | CD | NAG Subroutine Library(MARK8) |
NAP | Min | Tech | Circuit Design |
NEWPAC | Min | Tech | Finite Element |
OLYMPUS | Min | CD | Standardises Program design/structure |
OSDITTO | Std | CD | Tape and Disk Utilities |
OSFLOW | Std | CD | Flowcharting Program |
OPAL | MIN | HEP | OPAL-GEANT Monte Carlo Software |
PASCAL | Min | CD | Complier |
PATCHY | Std | HEP | Compiler Source code maintenance in multi-version programming |
PE2D | High | Tech | To Solve 2 Dimensional Equations of the Laplacian, Poissonian, Helmholtz or Diffusion Type |
PFORT | Min | Tech | Fortran Verifier |
PITFAL | High | CD | Location of Program Abend Addresses |
PLANT/SUPPLY | High | CD | Dynamic Substitution Facility for CMS |
PLUT078 | Std | CD | Molecular Drawing Program |
PL/1 (F) | Std | CD | Compiler |
PL/1 (optimizing) | Std | CD | Compiler |
PL360 | Min | CD | Compiler |
PPE | Std | CD | Problem Program Evaluation |
PREIFP | High | Tech | Interfaces FEMGEN and PACK |
PREVIEW | High | Tech | Interfaces FEMVIEW and PACK |
Std | SNS | Prints files on a GEC 4080 tape | |
PURL | Std | CD | Pre-Publication Text Preparation |
REDUCE | Min | CD | Algebraic Manipulation |
RL Library | Std | CD | Subroutine Library |
RL Utilities | Std | CD | RL Program Library Utilities |
SCRIPT | Std | CD | Text Processing System |
SIMULA | Min | ICF | Simulation |
SHELX | Min | DL | Crystallography |
SMART | Std | CD | Real Time Monitor for VM |
SMOG | High | CD | Graphics |
SPICE | Min | ICF | Circuit Design |
SRAM | Std | CD | CASORT Sort/Merge callable from high level languages |
SSP Library | Min | CD | Scientific Subroutine Package |
STACKER | Min | CD | 7 to 9 track Tape Conversion |
STAGE2 | Min | CD | Macro Processor |
STAIRS | Min | CD | Information Retrieval |
SUMX | Std | CD | Statistics Package |
TDMS | Std | CD | Disk and Tape Management |
TPELEC | Std | SMS | Copies 4080 tape files to ELECTRIC |
TRANFILE | Std | CD | File Transfer between CMS/OS & CERN,DESY,DL |
TRANSFER | Std | SNS | Copies ELECTRIC files to 4080 tapes |
TRIAB | Min | HEP | Book-keeping system for the analysis of experiments/tape management |
VICAR | Std | SA | Image Processing |
XEDIT | MAX | CD | Display Editing System for CMS |
XRAY74 | Min | DL | Crystallography |
ZBCOK | Std | HEP | A Dynamic Memory Management System |