SoftColl Meeting

8th European SoftColl Meeting

20th/21st, October 1994

The meeting has been held 20th/21st october 1994. Here is the Postscript final version of the agenda. A summary of the presented talks follows (minutes compiled A. Götz, ESRF).

Day 1 - 20/10/94

PS/CERN Presented their new control system architecture based on Unix and Lynx. They are replacing the old Norsk Data and CAMAC based systems. The whole of PS will be rejuvenated by Xmas 1994. The new system has a concept of workplace which is dedicated to one operator. Each workstation has two X11 screens attached to one workstation and one video screen attached to a 2nd workstation. There are 7 workplaces. IBM has been chosen to replace DEC (represents in total 100 workstations). The main problem in moving from DEC to IBM will be replacing the video capture units. ESRF A brief machine status: It is planned to have an uninterruptable power supply for the whole machine within May '95. Beam delivered time will reach 5000h in '95, and 6000h in '96. Lifetime is 60h at 100mA. The automatic machine realignment works fine (takes 20 min.) and is used every 3 months. Global orbit correction is perfomed every 5 min. The X-ray BPM was found unsuitable for local feedback. They're working on a superconducting ID and on a 7mm low gap vacuum chamber. Beamlines: There are 19 currently installed beamlines. The problem of collecting and archiving a huge amount of data from the beamline experiments has brought to a synergy of technology between the scientific and machine divisions. The resulting NICE project will use ATM technology for data trasfer from 100 up to 600 Mbit/s; commercial ATM interfaces are available both for HP and Sun. A Convex/Cray solution for general purpose processing is foreseen. A in-house board development using Analog Devices ADSP-2111 is in progress. BESSY Presented very briefly the old design for BESSY I (a standard model based system). BESSY II was presented in more detail. The highlights of the new system are: virtual device servers OOPS applications Oracle (or OODB) database C++ (+ C) and Ada (+ Fortran) Mapper and widgets to be maintained BESSY II api compatible ESRF api The lower level will be based on RPCs (device servers) which implement object oriented variable access (in Ada from ISIS). There will be an real-time (OO) database. NO operating system. CAN bus with Intel controller will be used. BESSY UIMS It is based on the Mapper/Formint pair (both implemented in Motif. The Mapper is an event-driven process which supports scheduling, non-blocking and a gui database. Formint is an action language interpeter with support for special variables for UI access and device access. The idea is to adapt the Mapper to the new BESSY II api. Forms will be in a Tex-like language. 90% of applications will be written in forms. Applications will talk to the Mapper and will not do their own X11/Motif programming. There are no limitations on the number of Mappers which can run. The Mapper represents 8-10 years of development (subcontracted to an external company, SPECS). A call from the operator to the equipment takes 100 ms. The goal is to achieve 10-15 ms. ELETTRA February 1992 - commissioned linac October 1993 - beam in storage ring November 1993 - first undulator light January 1994 - ramped power supplies first time Presently 3 frontends are installed. The goal is achieve 5000 hours operation per year. Installed today are 5 workstations, 1 server, 15 process computers and 70 equipment access units. Applications were available as of day one. Future developments are orbit position feedback, 2 more undulators, low gap vacuum chamber (17 mm) and a new rf gun for the linac. The linac control system has to be redesigned. The global position feedback is designed to damp beam vibrations up to 50 Hz by a factor of 2. It will be DSP based (probably 32 bit based). 84+84 correctors to control (vertical and horizontal). Control group is 9 people in total (hardware+software+networks). EPICS G.Surace presented the EPICS architecture and his impressions (based on the SSC EPICS environment) It supported two types of records - scan type and event type. Events were interrupts or software timers. There was only ONE software timer per IOC (VME frontend). This means events support was quite poor. No synchronous write was implemented. New record types can only be defined by the EPICS system group. EPICS does not use device drivers therefore it is likely that not all functionalities of a piece of hardware are implemented. EPICS does not have any synoptics support (it only has panels). PS/CERN Presented the MDR system. MDR is used for bringing large volumes of data from the low level to the upper level synchronously. Very useful for synchronous beam display. Supports coherent display from several sources. Uses a dynamic subscribe mechanism for applications. Scanning rates of up to 1 ms are supported. Uses RPC and UDP to communicate. It is adapted to frontend malfunctioning and workstation rebooting. Data is self-describing. This system is used for non-critical systems as well e.g. vacuum. CICERO K.Kostro presented the CICERO system. CICERO is an interface between different experiments and control systems. It is a framework for collaboration with the industry. It is object oriented. Supports data subscription. The key technology which is based on is CORBA. Presently it requires 920 kSF/year in investments. It is a kind of super-api or software bus. The sources are in the public domain. It will be used as a framework for the LHC. For more information contact J-M. Le Goff (CERN). CORBA Some words were said on CORBA to put it into context. An important point made was that CORBA is a STANDARD and not a product. Versions 1 and 2 will be interoperable. It is based on an object model. Static and dynamic invocation are supported. IBM has a CORBA implementation already called DSOM. CORBA objects inherit the class attributes in the interface definition language (IDL). Object persistancy is supported. BESSY II E.von Egan-Krieger presented the (new) device server concept which will be used at BESSY II. It is based on a distributed OO variable access method. There are 2 class concepts. A variable is be read/write. Variable belong to Variable Objects which belong to Property Classes which belong to Property Objects which belong to real device. Variables can be access points on the CAN bus. Device servers can poll a device.

Day 2 - 21/10/94

ELETTRA + GUI F.Potepan presented some of the new work going on on GUIs in Elettra. Users would like to see a ONE-BUTTON-MACHINE. The control group is working on this direction. It is foreseen to add a state notation language. Archiving needs to be done in a unified manner. WINGZ was not very well accepted by users. There was a need for more workspaces on the graphics screens. Hypertext is being used for online help. ELETTRA + error reporting C.Scafuri presented the error reporting system used at Elettra. Error reporting is done by the field access routines. In a first approach these errors were displayed in a dialog box. The 2nd approach was to define classes of errors - WARNING, TIMEOUT and NOT READY. Errors are put in a logfile for browsing at a later date. Now there is a need for filters to filter any unnecessary information. Errors have an error type and error number. BESSY II R. Müller presented an API interpreter. It will be very simple only reading standard input and writing to standard output. No flow control will be implemented. It will be used by users for "quick-and-dirties". Graphics Builders F. diMaio made a quick "Tour de Table" to find out which builders are presently being used. VUEIT from DEC is dying, PS is therefore looking to replace it. Builder Xaccessory is used at the ESRF. UIMX is used at Trieste + ESRF (EXPG). C.P.E. (home made) is used at Trieste. XFacemaker has been chosen by SL/CERN. World Wide Web G.Surace proposed that SOFTCOL has its own dedicated page in WWW. Everyone agreed that it was a good idea and that Elettra is the best suited to creating and maintaining this page. A question mark remained however to which server should be used. Turnkey systems G. von Egan-Krieger presented their solution on integrating turn-key solutions at BESSY II. The example presented was for an insertion device from Siemens. A device server will be written which communicates with the equipment via a real-time CAN bus. The device will appear as a black box. The user application will interrogate a database in the control room to demand authorization. The new gap values will be broadcast to all powersupplies via CAN bus. The powersupplies will have a lookup table with corrections to be applied. ELETTRA + Frontend hardware P.Michelini presented the experience at Elettra with frontend hardware. All VMEs at Elettra boot via tftp. OS9-NET file manager is used. Remote login to all systems is supported. Remote reset is possible from the control room. A command line interpreter allows any equipment device value to be readout or modified from the shell. Hardware problems experienced so far - HP disk, VME power supply, 4 x E6s, 5 x E16, 20 I/O boards, 2 Ethernet piggybacks, 2 RS232 piggybacks, 4 x MIL 1553 controllers. They have very few serial lines (some tens). On OS9 they have replaced the Microware version of select() with the public domain version. A simple call takes about 30 ms (time for RPC and MIL 1553 to execute). The synchronization system for the energy ramping is based on the broadcast mechanism of MIL 1553. Can work up to 1 KHz. ELETTRA - experience with LynxOS LynxOS is used for beamline control at Trieste. There is one central server for booting. Typical numbers of input/output channels/beamline are 2547/1601, 2399/1811 and 3803/2369 respectively. Development is done on a LynxOS system installed on a VME with remote disks mounted via NFS. Editing is done on Unix. LynxOS threads are used by control programs. All control is message based. Energy scans are done on LynxOS. Communication with user-based software running on LabView is via TCP/IP. WANF K.Kostro presented the West Area Neutron Facility. Control is done using a program called FactoryLink from USData. All programs are written in C++. An instance of a class corresponds to real world objects. Posix threads are used for multiple control. The system supports "update on change" (polling at 2 sec. and 15 sec. intervals). A "subscribe" mechanism also exists. This system will be used to upgrade the beamlines. The system supports CORBA compatibility. A paper on this can be found in the European Physics/Control WWW page. Elettra - user feedback beamline control S.Abrami presented the user's feedback on the beamline control at Trieste. The system supports multiple beamlines and frontends. User's may come with their own workstations. Synoptics are done with DataViews. An approach similar to the Elettra control room man-machine interface was followed. Drawings are imported from Autocad. Archiving of beamline parameters is supported. Xdesigner is used as GUI builder. The system has been working since one month. Local control of the experimental chamber is done with LabView running on a Macintosh and/or Unix station. Users advice to system programmers --go to graphical programming languages. Elettra - Photon BPMs The so-called PBPMs have been designed to work in "optimae conditiones operandi". They are based on the physical principle of photon emission. The present (prototype) is controlled with LabViews running on a Macintosh. The final system will run on LynxOS. The PBPMs are the first optical element in the frontend. They are designed to have micron sensitivity. Bending magnet contribution compensation is being done with DSPs. These sensors will be used for the future local feedback system. ESCATON A.Steindler presented the ESCATON energy analyser. Designed to deliver one spectrum every 50 ms and 100 values in a single shot. Readout will be done via IEEE. Has 96 anodes. Fibre optic connection to computer. LabView is used for control. Embedded computer software.

CONCLUSION

At the end of the meeting the following conclusions were drawn: Software Bus all participants agreed on the following prerequisites an equipment classification scheme, an identification scheme, and on using simple data types. all participants agreed that it is necessary to define a model, a root class and an api. it was recommended that a bottom up approach be adopted and to use an object oriented implementation. to produce a specification for an api and an object model within the next 6 month i.e. April 1995. SoftColl WWW home page it was decided to create a WWW home page dedicated to SoftColl Trieste very kindly volunteered to be in charge of setting the home page up and maintaining it. EC money Some of the participants were interested in applying to the European Community for money for developing software possibly within the context of SoftColl. All participants agreed to this. Institutes most interested were Trieste and Bessy. TRIESTE - big thankyou It was unanomously agreed to give a big thankyou to Trieste for their excellent seamless organisation of the 8th European SoftColl, for the hospitality and the excellent meal(s) they provided (the particpants will not forget the conference dinner in a hurry !).

Here are some photos from the meeting: