ILL - Operations

The operators, including card-punch assistants, loaded cards for batch jobs, and mounted tapes at users requests from the tape library. An important role was updating the archive each morning with incoming data so that it could be treated. System backup, and disk maintenance too was important. Initially users could use DECtapes for personal copies of important files. Later programs like TAPIP (Le Sourne, 1978) offered 9-track tape storage with the advantage of adding a directory to the contents at the start of the tape. This complemented system backup tools (FSKILL) used to clear off unused files.

In 1983 DEC abandoned the Jupiter project to build the next generation PDP10. There had been problems of synchronising the various CPU components; the future would hence be an emphasis on advancing the VAX systems. At the time of the acquisition of the KL1091S the performance of the VAX11/780 would have required 3 systems to replace it. By 1983 cluster software was becoming available which would simplify management of such multi-processor systems though clearly the overall power of the single cpu would not be available for intensive computing tasks.

With the acquisition of the DEC-VAX8600 the maintenance contract was renegotiated, allowing greater flexibility for the on-site technicians. The disk subsystems no longer required continued maintenance, and other components were generally much more reliable. This was a major element in DEC’s advantageous pricing for acquiring the new VAX processor. An HSC50 storage controller and a VAX11/751 (an upgrade of the former concentrator, PDP11/55) completed the basic VAXcluster.

The VMS system was very popular with the scientific community; programs could be shared easily with outside groups. The final closure of the DEC10 system was possible when the VAX8700 was added to the cluster.

The CISC instruction sets of mainframe computers required expensive look-ahead logic. On the VAX there were up to four or five locations addressed in each instruction, rendering it difficult to predict and optimise the use of the instruction and data bytes in the cache and number of cycles required in execution.. One example, "find first bit set and clear" for a 32-bit datum was only used by the operating system, but was in the deepest part of the job scheduler. The advent of the simpler RISC architectures lead to huge (apparent) cpu cycle rates. It is however interesting to note that today the Intel i86-64 CISC architecture has developed and maintains dominance commercially. Fortunately the overall power demands have been held in check by the micro-miniaturisation possible with modern nano-scale chip manufacture. Placing large caches within the cpu chip and using other second level cache techniques leave the user oblivious of the real advances in technology using the archaic i-386/pentium etc. instruction sets.

Later there was an addition of an AXP-3000 (romeo) to the cluster in 1990, running OpenVMS with the 21064 alpha RISC processor. This more than doubled the total processing power available. With the final acquisition of an AXP-3500 (alfa) in 1992 the 8600 and 8700 were taken out of service. The HSC50 disk subsystem was replaced by DECstorage towers of SCSI disk packs. One of the AXP systems was kept in working order until 2008, managed by the scientific computing group (RG), allowing the continued use of programs imported from external VMS systems. IP networking had been added to OpenVMS from the start. It also aided recovery of data from CDROMs containing the 36-bit data tape images from the DEC10 era stored in VMS-BACKUP files.

The Unix system was completely based on NFS networked file services which shared files amongst the mixture of HP, SGI, AXP-OSF, SUN, and even the AXP- VMS systems. The routine backups remained the responsibility of SI, services informatiques and progressively the system maintenance too for the variety of systems. In the course of time the HP and SGI were phased out, and linux on various PC platforms was adopted. This lead to increased uniformity across the ILL. Scientists still mostly used either Windows or Macintosh for their personal work. Few developed programs, though the gnu packages allowed both Windows and Macintosh to use similar code.

Initially there were two DEC GT40 Graphics Displays. These were PDP11/05 computers with a VT11 graphics board accessing the common memory. It offered scrolling ASCII terminal functions, but the VT11 could interpret autonomously move/draw commands to produce vector graphics on the screen. At the end of the display list a VT11 interrupt instruction allowed the main processor to rearrange the vector list before the display list was restarted. With long display lists the picture started to fade before being refreshed, limiting the total number of vectors displayed. The first general purpose graphics routines were written for this device (using the GT40A library, then BOOK, E Douieb) with the now familiar set of four x by four y axes which became a characteristic of ILL displays. The light-pen was not very useful as a control device for simple plots, mostly being used to control the legendary moon- landing program.

The Calcomp pen plotter was heavily used, but was very slow and not ideal for the many data points coming from ILL instruments.

Advent of the Tektronix 4010 storage tube terminal with attached silvered-paper hardcopy finally gave the scientists an interactive terminal with a useful hardcopy output record. The X-Y cursors allowed selection of peaks which greatly aided data analysis. After two had been installed on the central computer they were also progressively added to the instrument; the hardcopies were shared. In operation the electron beam was positioned and moved on the screen. Secondary electrons were picked off an adjacent grid, and used to rewrite the screen. The two beams were gated to operate independently. Erasure of the screen was by flooding the whole screen and the switching off the rewrite temporarily. The screen had a limited brightness, and was subject to burn-in and fade with use. The basic library PLOT10 (and simpler vector code generators) was available for many systems; it was quite easy to generate the tagged characters which represented the vectors and send them through terminal handlers even on CDC-Cyber systems, and other non byte machines.

In later years the basic PLOT10 MOVABS/DRAWABS etc., commands were interpreted with emulations on raster graphics systems which were much more comfortable to use, though actually lacked the precision of the storage tube with its pure vector graphics. One example was the Hercules graphics board, common on early scientific IBM-PCs. At the ILL the Pericom terminals became a standard, though quite cheap emulator terminals, e.g. the FALCO with a single chip processor, were popular amongst scientists.

A Versatek off-line electrostatic printer took over much of the Calcomp plotting. The plot files were spooled to disk, then copied at intervals to a tape which drove the plotter off-line with its plain paper output.

One VS11 16 colour graphics unit was acquired for D19, but proved inflexible to program. For the small angle scattering instruments one ATARI-1040ST program was written (Ghosh, 1985) which took serial ASCII data and displayed one or four 64x64 spectra in 16 colours (display using 600x200 pixels). This ATARI microcomputer had a graphics chip with a colour lookup table, making possible a wide range of colour representations.

Over 10 ATARIs with associated SONY display monitors were connected by serial lines both on instruments and the central computer. The display driving programs generating the ascii serial data could use a shared memory segment which was available both on the instrument computers and the central VAX computers. Some were used for monitoring incoming data, others for treated data on D11, D17 and D16. 35mm cameras could be attached to screen-mounts and using polaroid colour films these yielded transparencies for subsequent projection of results at modest cost.

Benson colour pen plotters were also much used, especially for crystallographic structures and gamma-ray level schemes.
Graphics software was primarily based on PLOT10 and the pen-plotter standard libraries. Computer Associates DISSPLA had some usage, but programs were not transportable.

From 1988 VAXstations with graphics capabilities were installed on instruments. On D7 much use was made of the VAXstation-II GPX graphics (Schaerpf) (pre X- window) using Basic. The X-window graphics were not easy to program, lacking documentation. At this stage a commercial package by PVwave was acquired, and this lead to a rapid development of display software using the IDL language. Later as PVwave showed less interest in scientific display contracts were established with IDL, who had meanwhile developed their own graphics capabilities for these systems. The cost of licenses remains a significant recurrent cost, though it is possible to create exportable programs , (LAMP, Richard, COSMOS, Cubitt, etc), containing a free run-time IDL kernel.

For displaying large biological molecules an Evans and Sutherland PS340 display system was acquired (1988) and used primarily by the biologists. This model, with a raster display ran autonomously from the PS2 vector display which was usually integrated into a VAX system. Typical programs used included FRODO for rearranging parts of molecular strands.

In March 1985, the Apple LaserWriter was the first printer to ship with PostScript, sparking the desktop publishing (DTP) revolution. It was now possible to insert nice plots in a printed document without the help of a pen plotter, scissors and glue. The ILL had a Macintosh AppleTalk network that allowed users share such documents and print them.

The first A4 Postscript color printer arrived at the ILL in 1994, a Tektronix Phaser II SDX 4648 using a sublimation technologiy which provided high quality images. Then a Tektronix Phaser 340, a solid ink color printer, was bought in 1996. 

At the ILL. graphical output as PostScript files became increasingly common after the introduction of general networked printers. Poster printing and embedding EPS files in documents became common, helped by the initial acquisition in 1997 (with contributions from the ESRF and EMBL) of a Tektronix Phaser 600, a solid ink A0 Postscript printer, later replaced by HP PostScript A0 printers. The cost of this type of printer rapidly dropped, and PC/ Macintosh software (Aldus PageMaker, QuarkXpress, Adobe Illustrator, later PowerPoint, etc) evolved to allow easy creation of posters of scientific results. Networked PostScript laser printers too became common on instruments for graphics as an alternative to screen dumps from terminals.

Initially the DEC10 console computer was a KSR35 heavy duty teletype (10cps), though this was soon changed for a DECwriter LA30 80 column matrix printer (30cps). A VT05 VDU terminal offered dynamic visualisation for controlling processes. The users initially used 6-8 Teletype ASR33  (electromechanical 80 character 10cps cylindrical print-head) and two DEC GT40 displays. The teletypes were located around the ILL, each connected with a current loop circuit.

The introduction of two Tektronix 4010 storage scope graphics terminals (13 inch) and their 4610 hardcopy (1978) lead to rapid development of interactive graphics for viewing and analysing data with x-y line plots, complemented by 2D contour plots of multidetector data. The Diffraction group actually financed one of these terminals (Alan Hewat) and had priority of use during part of the day. Refinement of powder diffraction data was a production activity (Hewat). Because true vectors were drawn the diffraction patterns were of much better quality than those later produced on raster screens.

From 1978 the DECwriter II, the LA36 9 pin 132 column printing terminal (30cps) was being introduced. Most were subsequently upgraded with a third party electronics board (FUNGUS) which could print at 120 cps, with bi-directional optimisation. Though not capable of true descender characters it could print upper and lower case.

When the DEC1091S was acquired additional DEC VT100 vdu terminals, and later Plessey PT100 compatible vdus were added to the park. A small number of Tektronix 4025 raster (640x350) graphics terminals were also acquired, though the resolution was inferior to the storage screen terminals; the scrolling alphanumeric capabilities and bright displays were quite welcome. We discovered that the operating system was actually in upper and lowercase for the first time; terminals had full character sets, and the original LP10 drum printers were complemented by two 1200 lpm LP20 charaband printers with full characters sets.

The formatting program RUNOFF was used for documentation with a daisy-wheel printer for improved quality. DEC later produced the DECwriter III, the LA120, which printed at 120 cps, and became standard as console printers on instruments. The VAX11/730 computers were delivered including LA100 console terminals (120cps, desktop 120 character wide printers).

The Pericom Grafpak (13 inch 1024x800) ), then the Pericom Monterey MG200 (17 inch, 1986) were the first modest cost raster graphics terminals which had widespread use. Later the much cheaper single chip Falco (13 inch) was cheap enough for scientists to have graphics terminals on their desks.

Matrix printers could print screen copies. All these terminals used the Tektronix PLOT10 ASCII print coded graphics, connected via serial lines (9600 baud).

The terminal network was extended progressively throughout the ILL from 1982 with the introduction of a MICOM, (later Satelcom) terminal switch; this allowed access from any terminal to either the central system, or the instruments. One terminal on each instrument also could use this to access the central computer.

Following the introduction of unix, X-window terminals were added to the network; there was less need to augment the number of workstations. The cost of the 19-20" CRT colour monitors, the principal component dropped rapidly. The advantages of a non force ventilated unit in offices was appreciated. This again changed with the advent of flat screen TFT monitors which liberated considerable desk-space taken by the CRT screens.
The X-window server in Cygwin PC software and the gcc compilers enabled building versions of unix software to run on PC-Windows. The large LAUE package from John Campbell at Daresbury (40000 lines of Fortran 75000 lines of C) were slightly modified for use on Windows. Some scientists used commercial X-window servers on PCs (Xserve). With OS X, Macintosh computers had XQuartz X-window server software built-in for using the unix graphics.