After a long period of uncertainty about future equipment and systems things are now beginning to happen. Users are likely to see several changes during the next 18 months to two years. Efforts within Computing Division are directed towards a continuity of service consistent with taking advantage of the new hardware and software that will be acquired.
The RAL Computer User Liaison Committee met on April 30th and June 4th. Among other items, it requested that each Representatives Meeting examine its Terms of Reference.
FORUM 25 carried an informative article on various stages of replacement of the 195s by the IBM 3081D. It is believed that the effect of this on users will be slight, with some changes in /*NEEDS control cards, short interruptions in service, some differences in elapsed times and, later, increased memory availability.
Initially a scale factor of 0.8 will be used for accounting cpu hours on the 3081. This will be reviewed at approximately quarterly intervals during the current allocation year.
There were major problems with power on one of the 195s after the air-conditioning maintenance in April. Approximately 95 hours were lost on 195/1 and 55 hours on 195/2. A director fault on the 3032 caused a down time of 3.5 hours. This was most probably caused by air-conditioning and alternator problems which resulted in a total of 10 hours lost time. A major fault on 195/2 was responsible for lost time of approximately 44 hours in the week ending 20 June. Batch turnaround was affected by all of these faults.
Problems with fixed disks necessitated restores of both ULIB03 (after air-conditioning shutdown) and RHEL01 (week ending 23 May). In both cases it required an HDA replacement to the drives.
Following a head crash on one of the MEMOREX 3350 equivalent disk drives on Friday 16 April it was necessary to restore all user libraries (ULIB's) from back-ups taken on the night of Wednesday 14 April.
It is intended to run a courier service between R27 and the stores pickup and delivery points in R2, R3 and R25. This will commence as soon as suitable collection boxes have been arranged.
There are three new terminal pools in the Atlas Centre. The graphics pool is in R31 051/52, the Starlink terminal pool is in R30 3/4 and there is a new pool in R26 West with IBM 3270 compatible terminals and documentation printers. The pool in R26 East no longer exists.
Users are again reminded that anyone who believes they will need continued use of the central computer's 7-track magnetic tape drive should contact P C Thompson as soon as possible. It is intended to withdraw this facility on April 1st 1983.
Datasets created since 1st May 1982 using VOL=REF=FREE were irretrievably lost early on Wednesday 9th June due to operational problems. Steps have been taken to safeguard against this happening again in the future.
The charge factors for ELECTRIC and CMS, introduced at the beginning of the financial year, are as published in the March issue of FORUM (21). The maximum charge rate for jobs run in CMSBATCH is 0.4. Overdrawn users may login to CMS or ELECTRIC after 1600 on Fridays.
On 21 June the 10 hours allocation of Priority 1 time to each account was withdrawn. Users are now issued with a PI allocation equal to the total weekly allocation, except perhaps in those cases where the allocations are in the control of the Category Representative. The sum of allocations at other priority levels remains at 1.5 times this value. Users experiencing difficulty with their- P1 allocation should contact their Category Representative if they have one, otherwise Mrs Sue Ward, CMS USERID=SHW, ELECTRIC ID=FE.
Category Representatives are now able, each week, to list by account the usage of time at all PRIORITY levels and of CMS and ELECTRIC AU's. A similar facility will be provided for Account Representatives at the ID level but this will not initially include information on CMS and ELECTRIC usage.
As ELECTRIC is being phased out, no work other than normal maintenance has been done on it. A plan is under preparation for the run-down of ELECTRIC.
There has been considerable progress towards two ISO (International Standards Organisation) graphics standards: one for graphics systems and one for graphics file formats. RAL Graphics Section took the decision earlier this year to implement GKS, the proposed ISO graphics system. This decision now appears even more sensible as GKS has been ratified as a Draft International Standard and six countries (UK, USA, Germany, France, Netherlands, Japan) have indicated their intention to adopt GKS, without change, as a national standard as well. The importance of the USA being in this group is obvious, considering the proportion of graphics equipment made there. For graphics metafiles, a joint USA/ISO effort aims to define a metafile standard by about January 1983 and bring the standard to the same status as GKS now has by May 1983. RAL staff are involved in these discussions and support of the metafile standard will be provided when its format has been agreed.
Users will have noticed that part of the RAL Graphics Manual has been circulated. Many more chapters have been typed and will be distributed when the illustrations and examples have been produced and checked.
Support for all FR80 cameras and all Sigma terminals has been written and installed under CMS as well as experimental support for the Calcomp 81 desktop A3 plotter. Work is now in hand on facilities for high quality text, on the FR80 and on less capable terminals. As part of this development a large library of fonts is being organized.
There will be no support for seven track tapes. Nine track tapes will be supported at 800 bpi and 1600 bpi. The maximum blocksize will be 1536. Spanning of FR80 files across multiple tapes will be supported. The facilities provided for use under MVT will need some modification for MVS. No support will be provided for CMS. Multiple files will be allowed. Details of how to use the facilities will be published later.
As stated at the previous meeting, the general implementation of the MVS operating system will have three main phases. The first phase will start after the 3081 installation is complete, when a few basic facilities will be established. The second phase will allow a limited service to some users. During this phase components required for the full service will be implemented. This test system should be ready approximately a year after the 3081 has been running effectively. The production system is expected to be operational after a further period of about 10-12 months. The ATLAS-10 (Fujitsu M380), if available, will be integrated into the production MVS/JES3 complex. The third phase will involve the full system when users will be migrated from the MVT service. Firm timescales will be provided as soon as the 3081 is fully operational.
The present auto-archive library management system involving the ULIB's cannot be implemented for MVS in a satisfactory manner. Alternative arrangements are still being sought.
Please contact PAO if you think your program may be MVT or HASP dependent so that they can assess the level of program modification required and help you to be ready for MVS when it comes.
Users who submit jobs for tapes which are not in the local library may, in certain circumstances, notice the job status fluctuate between HELD-WAITING TAPES and not held. This is caused by the design of TDMS whereby all of the jobs which are HELD-WAITING TAPES are released when the librarian does a TDMS update. Jobs whose tapes have not been moved into the local library only become held again when they reach a sufficiently high point in the job queue. How to overcome this is a subject of study.
In order to overcome problems with file transfer mechanisms using pseudo punches, which need to transfer more than 9999 card images, a specification of a cards limit of 9876 is interpreted as a very large number. The new maximum is unlikely to be reached.
VM Spool - Users should note that very old spool files are subject to deletion without notice. This is because the VM spool can become full, thus causing operational problems.
New CP and CMS Versions - The present version of CP is 1.7, and of CMS is also 1.7 though version 1.12 of CMS is currently under test.
The following is a very brief summary of the main changes to CMS since the last User Representative Meeting. Further details can be found in the NEWS files and the HELP system. The main changes since the last CCRM were the introduction of XPLANT and the FREEDA storage system.
No change has been made to any of the Fortran compilers since the last meeting. Current releases are as follows:
Release 1.1.1 of the VS Fortran compiler and its library have been installed under a trial version of MVS and found to run satisfactorily (limited testing only has so far been carried out). CMS users are advised to start making use of this compiler (the current release is considered stable). The new library VFORTLIB (CMS TXTLIB) will support files compiled under FORTHX or FORTGI as well as FORTVS.
IBM have announced release 2.0 of VS Fortran which will contain bit-handling functions (similar to Fortran H (Extended Plus) compiler functions), and language extensions for Hollerith and hexadecimal in DATA statements. A considerable programme of enhancements is envisaged for this compiler over the next few years (including the possible addition of SLAC traceback code - MR CLEAN).
The IBM manual VS Fortran Application Program: Language Reference (ref no:- GC26-3986-1) may be ordered from the RAL Computing Division Documentation Office.
Future Policy - The H Extended Plus compiler produces object code which is primarily intended to run on a 360/195. The special coding concerns instruction scheduling which is performed during the level 3 optimization. It is based on a standard IBM compiler known as H Extended that just has optimization levels 1 and 2. Because 195s are so rare, H Extended Plus is very expensive to rent and difficult to support. Ultimately we expect a Fortran 77 Compiler to be the main one used here. In the meantime it might be sensible to exchange H Extended Plus for H Extended. The manuals etc are the same, and language extensions are compatible. With the departure of the 195s the expense seems no longer justified.
Two new bugs have been identified in the VS Fortran compiler release 1.1.1 - fixes for them are awaited from IBM. One bug appears in code that contains an expression for an I/O statement inside a DO-loop when the DO-loop controlling variable is used in the expression. This bug appears at optimisation levels 2 and 3 only. The second bug concerns bad source code whereby a statement WRITE(1) IWORD(2:) with IWORD not dimensioned will cause the compiler to ABEND.
One bug in the Fortran H (Extended Plus) compiler is being investigated whereby the statement W = 1.0Q0/2.OQO (extended precision - REAL*16) causes a compiler ABEND at optimisation levels 2 and 3- This problem appears under CMS only.
SYMBUG is an interactive debugging facility provided for1 RAL by Computer Associates Incorporated as a rapid debugging aid for use with Fortran programs. At the moment it is under test, but when it it is fully implemented, it will replace the IBM Fortran Interactive Debugging Aid, FIDBUG (currently on the U-DISK), with a fully supported and much enhanced utility. Additional features of SYMBUG include command abbreviations, a small HELP system, CP and CMS commands, and CMS DEBUG facilities.
A significant number of private and public 3330 disks have already been withdrawn. Nine of the busiest 3330s have now been permanently mounted. All of the public 3330 disks, except for one system disk still to be moved, have been transferred to 3350s. This has had an impact on the number of disk mounts.
A number of users still have to be contacted so that arrangements can be made for data on these private disks which are suitable for withdrawal. Busy datasets should be moved onto public space. Idle datasets should be deleted or written to tape. It is an operational goal to eliminate disk mounts.
It should be appreciated by users that the use of O.S. datasets is much easier if they are accessed through the extensive cataloguing system on MVT. Users are therefore reminded that, as suggested in the notes for the previous meeting, datasets that are not catalogued do not conform to the conventions adopted at RAL and as such will shortly he subject to deletion by the housekeeping programs.
Sixteen workstations have been converted to VNET so far. The first conversions were on the RAL site and brought to light some problems which system testing could not produce. The major problems have been solved and the conversion programme resumed.
Most of the networked GEC2050's not due for replacement have been, or are about to be, converted. During July and August it is intended to convert some more networked machines that are not GEC machines. The GECHOOO machines with filestores will be converted when an appropriate new release of the operating system is ready in the near future. Development work is also necessary for the SERC Primes. About 14 HASP workstations will not be converted to VNET. It is intended to have a provisional schedule ready for the date of the meeting. It will be maintained in a CMS HELP file, visible by typing 'HELP VNET SCHEDULE', A list of outstanding problems will be kept in CMS, visible by typing 'HELP VNET PROBLEMS'.
Several HASP workstations are due to be changed or superseded. Some GEC2050's are due to be replaced by GEC4OOO's, or have their users transferred to SERC MUM's in the same university. The continued use of a number of connections is still under consideration. The aim is to substantially complete the transition to VNET before the end of this calendar year.
The chief development by Telecommunications is the introduction of the new Packet Switching Exchanges, which are now located at ERCC, ULCC, Cambridge and CERN. A further machine is about to be installed at Newcastle. This has enabled the connection :f various new hosts and workstations, and a programme of rationalisation aimed at cost saving, is new under way.
The first area for rationalisation is the connection of existing hosts to their nearest F3E. The second aim is to convert from the old communications protocol (called binary synchronous - a half-duplex method) to the full-duplex, high-level Data Link Control (HDLC). The third area of rationalisation is to share the load on the RAL PSE between two machines, RLE1 and RLE2. The opportunity is being taken to move the machines of individual families, such as the ICF GEC's and ICF Primes, which connect to the Network at RAL, to be connected to the same PSE's.
With such a program of work, there is inevitably some disruption to the Network from time to time to enable exchanges to be reconfigured, etc. The effects of this, however, are being minimised by restricting such changes to Tuesday mornings (0830 - 0900) with test sessions for new exchange software between 1730-1930 on Thursday evenings. Warning of any changes will be entered in the NETSTAT machine at least one day in advance.
An experimental line has been installed between the 3032 Front End at RAL and the IBM 3081 at CERN. Tests are being carried out using the IBM supplied Network Job Entry (NJE) system. Jobs submitted from RAL, normally by logged in CMS users, run at CERN and output received back at RAL. Similar tests are being carried out in the opposite direction. When the basic system has been proved, a similar experimental line will be installed to DESY. It will then be necessary tc identify what facilities might be needed in order to provide services encompassing VNET workstations based at RAL and RIOS (Remote Input Output Stations) stations elsewhere. It is expected that this work will enable the existing back to back connections to be removed.
A new version of the PACX software will be introduced on Monday 2 August. The important change is that services may then be selected by alphanumeric names, eg CMS, RLGB. Speed selection will be done automatically. However, if a terminal set at, say, 4800 baud finds that there is a queue for the service, the user will have to change the terminal speed before attempting the same service at another speed. PACX will recognise SERC network names, plus CMS, ELEC, and CERN. The existing numerical system will no longer be appropriate in most cases.
It has been found that some workstations do not have channel 8 on the carriage control tape set correctly. This is used by HASP to print a line of asterisks along the perforations between pages at job start and end. If the line of asterisks is not on the page perforation the control tape is incorrect or the sensing mechanism is wrong. It is important that such faults are rectified, because VNET banner pages use channel 8 and if that is incorrectly set then VNET banner pages will not be properly separated from the tail page of the preceding output.
The ROUTES routine has been modified to return the secondary route parameters if supplied with a second argument. It is upwards compatible with the old version. Several other routines from the SY chapter have now been made available on CMS. They are MVTID, CLOCK, SPM, GPM, CLFILE, STOP, KLOCK and DATE. A new routine, INDUSE, is also available on CMS. It returns the values provided by the CP INDICATE USER command. It is not available on MVT. See 'NEWS RHELIB' for further details.
The new CERN library is now available. Users on CMS should type 'NEWS CERNLIB' for more details.
There have been a number of problems with the new library because there are several routines which appear in both RHELIB and CERNLIB, yet have slightly different specifications. In order to remove these inconsistencies, it is proposed that on 11th August the following will be done:
After three months, UZERO, UBLANK and UFILL will be altered so that they will abend if they are not called with 3, 3 and 4 arguments respectively. After a further three months, UZERO, VZERO, UFILL, VFILL, UBLANK and VBLANK will be replaced by the versions from CERN.
The question of support for the DRIO package has been resolved. It will continue to be supported under MVT and MVS and a CMS simulation will be provided.
++U - This facility for communication with the PAO has now been removed. Users still have use of MESSAGE (ELECTRIC), ASKUS, TELLUS and GRIPE (CMS), and the telephone Abingdon 21900, ext 6111 if they experience problems. It should be noted that an 'Ansafone' is attached, so that messages are recorded out of office hours and when the advisors are busy.
++T - This facility has also been removed. Messages to the tape librarian should be sent to ELECTRIC ID=JU or to the CMS userid=TAPELIB.
Screen Terminals - Building R1 now has four IBM-3278 terminals for use by general users. A document is now available describing the elementary principles of their use and indicating how to find out how to use them to great effect. Users having difficulty using them should get in touch with PAO.
CIFER Terminals - An enhancement to the CIFER terminal to enable it to emulate IBM full screen working over the network has been designed and a prototype is expected in a few weeks. If this is successful it should be possible to convert some of the existing CIFER terminals eventually.
TDMS Write-up - This has been produced as CIGAR section D12 and distributed with FORUM 25.
TDMSUSER - A catalogued procedure called TDMSSORT has been installed on the MVT Batch system. This procedure performs all the functions provided by the existing TDMSUSER procedure but in addition the lists of volume serial numbers generated by the LIST function will be sorted in alphabetic then numerical order. The use of TDMSUSER functions is described in CIGAR section D12.
The next meeting of the DECsystem-10 Users Committee will be held on Wednesday 1 September 1982 at 10.30 am in room 5210 of the James Clerk Maxwell Building, King's Buildings, Edinburgh. The new chairman is:
Miss Anne MacKinnon
Department of Electronic and Electrical Engineering
Rankine Building
Glasgow University
GLASGOW G12 8QQ
Tel: 041 339 0694X8855, Ext 7232
and the Secretary is:
R J Hare
SERC DECsystem-10
ERCC
King's Buildings
Mayfield Road
EDINBURGH EH9 3JZ
Tel: 031 667 1081, Ext 2619
Routine Preventative Maintenance will take place on the following days from 1800 - 2200 hours. Login messages on the ELECTRIC and CMS services will be issued prior to each maintenance session as a reminder.
The next shutdown of all computer systems (except network equipment) scheduled for 1982 for -the maintenance of air-conditioning plant is:
0800 hrs on Friday 22 Oct till late Monday 25 Oct
System development is scheduled on Wednesday mornings from 08.30 to 10.30 and Thursday evenings from 17.30 to 19.30.
