This last period before the end of the Financial Year has seen an overall satisfactory performance from all machines and a continuing pattern of low demand for resources. Apart from odd fluctuations users have not put any large demands on the facilities. Problems still occur associated with the storage overwrites on the IBM 3081 but the frequency has dropped significantly. The Prime upgrade program is almost complete and has been carried out more or less according to plan. It is pleasing to note that the new P9950 at UMIST is running well and measures up to its advertised power of 2.5 that of the P750. Full screen editing facilities are now available on the Primes with the installation of the EMACS package. PROLOG has also been mounted on these machines.
An important meeting of the GEC User Group discussed and agreed the effects of the reduction in manpower imposed by the Central Computing Committee. In particular there was unanimous agreement to freeze the OS4000 Operating System at the present release level. A full report of this meeting is in this issue. SERC and the COMPUTER BOARD held a successful Regional meeting in Edinburgh on 31 January which is also reported this time. Another such meeting covering Wales and the South-West is being held on 28 June in Bristol. Fuller details are included in this FORUM84.
The RAL Computing Division has for a number of years provided user education through various courses given by its staff. We currently face serious problems in trying to keep these courses going as described later in this issue.
The Program Advisory Office is also being studied to try to restore a better service for the IBM users. This is being looked at as a Divisional responsibility involving all of the relevant groups. The outcome of this should be publicised in the next few weeks.
The policy on the magnetic tape archive will soon take effect. All users with tapes in the archive must read the final warning notice contained in this issue.
In 1979, the Computer Board approved plans for replacing ULCC's CDC mainframes and designated it as a national centre. The growth of the CDC service in the 1970s had been accompanied by the establishment of a network predominately for batch traffic with lines between ULCC and remote job entry stations (or emulators) in London and at other universities. The 1979 plans included the provision of completely new communications arrangements. Packet switched networking facilities were to be developed to handle traffic among all sites connected to ULCC. Access to ULCC's new mainframes was to be based on the high level protocols adopted as standards by the UK academic community. These are X29 for terminal access, FTP for file transfers and JTMP for batch traffic.
The diagram below shows the communications arrangements being implemented at ULCC and their relationship with the national networks. To emphasise the main features, some of the details have been omitted. The next sections deal with the various components and their present status.
The switching facilities at ULCC are provided by a number of interlinked modular systems. The equipment constitutes a network in its own right with an initial total of over 70 links as well as connections to JANET and PSS. Automatic monitoring and management facilities are provided including real-time operator displays of components status and call totals.
This network serves not only to switch X25 traffic among sites connected to ULCC but also to concentrate traffic destined for ULCC's mainframes and to provide paths to the national networks. The service became operational in October 1982, the main initial use being for X29 terminal traffic to various hosts.
The original intention was that both batch and interactive access to the Amdahl should be via a high-bandwidth path from the switch. In the event, technical difficulties caused significant delays in the development of support for an adequate number of network terminal sessions on the Amdahl. In the meantime, appropriate IBM software had become available to provide such facilities. Therefore, to avoid yet further delay, an Amdahl 4705E communications processor and the associated software were installed at ULCC in December 1983 giving a separate path for interactive traffic as shown in the diagram. An initial service supporting up to 64 simultaneous sessions was announced at the end of January 1984. The intention is to increase this figure during the next few months to 120; the original target in the centre's plans.
From the outset it was decided that the batch access to the Amdahl/Cray should be by means of JTMP; the agreed standard Job Transfer and Manipulation Protocol. This is to be the only batch protocol supported by Amdahl. ULCC's implementation is based on a portable package being developed at the University of Salford. The magnitude of Salford's task and the extent of the work required to install the code at ULCC were significantly underestimated. However, a trial service version has been operating since September 1983. Exeter is shown on the diagram because it was the first site to offer a user service based on JTMP to ULCC via JANET.
The latest release of the Salford package containing major improvements has been installed at ULCC and is the new service version. ULCC's present mechanism for shipping JTMP data is an interim protocol. A version using FTP as the carrier is being developed in conformity with community standards. It is hoped that this will reach the testing stage by April.
Most of the ULCC's user sites run equipment using old proprietary protocols such as HASP or UT200. Thus it is necessary for protocol conversion to be provided between such protocols and JTMP, as shown on the diagram. In time all sites will have JTMP capabilities of their own and their connections will then be directly to the switch with no protocol converters on the route. A protocol converter at Daresbury provides translation between the standard protocols supported by ULCC (X29 and JTMP) and those used by some of the SERC and NERC systems (ITP and networked HASP).
Protocol converters for the UT200 and HASP lines to ULCC are undergoing extensive testing. There are about 60 links whose traffic requires such conversion and the resultant JTMP load will be considerable. The process of transferring the connections on to the converters is expected to start in the summer.
At the end of January, a trial service was opened between Imperial College and ULCC based on protocol conversion at ULCC. QMC and Cambridge are expected to transfer their existing connections onto the same route by the end of February.
A service has been in operation since September 1983 to provide access to ULCC for SERC/NERC users via JANET and the Daresbury protocol converter. Although there have been deficiencies in the initial version of JTMP which have caused problems for users, the more serious bugs have now been removed. A significant improvement in performance is expected when the service is moved onto the latest release of the Salford package in March.
The developments described are part of an ambitious programme to accomplish a transition from ULCC's original RJE network based on CDC mainframes to a modern communication scheme using standard high-level protocols. The main components of the new arrangements are now nearing completion. The coming months will see the introduction of further services and improvements to existing ones. Also during this period, the transfer of the remaining connections from the CDC world onto the new communications scheme will be undertaken.
Future areas of work will include the provision of a free-standing FTP and the implementation of advanced JTMP facilities, especially in the area of job visibility.
A meeting was held in London on 1 March which was well attended.
The Configuration section of the database now contains hardware details provided by the members.
The meeting expressed interest in holding a joint meeting with the Computer Board VAX Groups with a view to establishing a central database for VAX hardware and software.
In the afternoon a presentation by representatives from DEC on their future hardware and software plans was followed by discussion.
Users are reminded that they must contact the Tape Librarian in order to save their tapes that have not been accessed for the last five years from being scrapped. A list of these tapes was published with the last issue of FORUM and a copy of this is available on CMS.
A list of the tapes not identified by 30 April 1984 will be re-issued in CMS. Thereafter, tapes not identified by 31 July 1984 will be removed from the Library and scrapped!
This was held at the Rutherford Appleton Laboratory on Thursday 27 January 1984. There were 34 attendees and most sites were represented.
It was a special meeting called to discuss the future support of the GEC 4000 machines in the light of decreased manpower caused by reduced funding. A paper produced by Mike Jane, Head of User Support Group, gave the options.
Professor Bob Hopgood, Head of Computing Division, outlined the spending reductions required over the next five years. These have been made necessary due to funding cuts from some of the Boards but also due to unfavourable exchange rates. Support for GEC 4000 machines will be reduced from the current 5 man years/year to 4 man years/year; some lightly used and obsolete machines will be removed.
After consideration of Mike Jane's paper a vote was taken on which of the options should be adopted. The result was unanimously in favour of a freeze in the operating system together with support from RAL being maintained, albeit at a reduced level. Some of the development work will be possible and users were invited to express their views on such development activities.
Questions and discussion highlighted items of importance; these were voted on to give some idea of priority:
Release 3 of CMS will be installed during System Development on 4 April. It is upwards compatible with Release 2, except for the LISTFILE command, which users have already been warned about.
The major change in Release 3 is the introduction of the System Product Interpreter (also known as the Restructured Extended Executor, or simply REXX). The REXX language is totally different from EXEC and EXEC-2 and is similar in general appearance to PL/l. It provides programming constructs such as IF...THEN...ELSE and DO...END, subroutines, internal and external functions and, should you need it, arbitrary precision arithmetic. A complex EXEC written in REXX will usually be much shorter than the same EXEC written in EXEC or EXEC2 and will be correspondingly more efficient. EXECs and XEDIT macros written in REXX are identified by starting with a comment line of the form:
/* ... any comment ... */
REXX is described in a reference manual and a User guide. These are on order and should be available from the documentation officer by the end of March.
Changes specific to fullscreen terminals include the ability to split the screen in an arbitrary fashion and prefix command macros.
All of the Rutherford written EXECs have been modified to replace obsolete commands by IBM-supported commands, wherever possible. CPEX, EXEC EMSG and STACKF have all been replaced by EXECIO. All QSET variables with names beginning SY have been replaced by GLOBALV variables with the same name in the group *RALSYS. For example, the information previously held in the GLOBALV variable SYJBCARD will now be held in the GLOBALV variable SYJBCARD in group *RALSYS. The EXECs have also been modified to replace LISTFILE by LFILE where necessary.
The OPTIONS and CFILE commands run as nucleus extensions, to improve overall efficiency. The VPRINT command has been modified to allow reader files to be transferred to a VNET workstation. The SPSAVE and SPREST commands have been enhanced to allow nested calls.
Over the summer. 15 lower sixth form students from 3 local schools each spent three weeks working in the Computing Division. The intention was that they should gain experience of working in three separate areas, one being Operations. In the event, two were put on to individual three week projects while the others continued as originally intended.
The pupils found that the programme that had been arranged for them was rather different from their prior expectations but nonetheless the experience was felt to be very valuable. Apart, from the period on Operations, work included text copying, editing and verifying on the G-EXEC help system, transferring a database from PRIME to IBM machines, writing options for the Telecoms termfaults database interrogation program and also occasional other tasks such as unpacking and testing terminals and entering data into the Telecoms info database about SERCnet. There were two individual three week projects arranged - one was to provide a modification to NETLINK to allow mnemonics to be used instead of the ID code, the other consisted of the building and programming of a micro to drive Speech Board.
During the week on operations the pupils had the benefit of a couple of teaching sessions introducing the work or the Telecoms section, an area that is hardly touched in the schools.
The exercise was certainly very valuable and we look forward to something similar next year, although the format adopted next time will almost certainly be rather different, drawing on experience of this year's pattern.
