- Email: [email protected]
The rejuvenation status of TRISTAN accelerator control system
Nuclear Instruments and Methods in Physics Research A 352 (1994) 128-130 North-Holland
NUCLEAR INSTRUMENTS & METHODS IN PHYSICS RESEARCH Sect,onA
The rejuvenation status of TRISTAN accelerator control system T. M i m a s h i a, A. A k i y a m a a, S. Araki a, K. K u d o h ~, I. K o m a d a ~, T. K a w a m o t o a, S. K u r o k a w a ", T. N a i t o h a, S. T a k e d a a, j. U r a k a w a ~, T. T a k a s h i m a a, K. F u r u k a w a ~, J. Navratil b K. O i d e a N. Y a m a m o t o a a National Laboratory for High Energy Physics (KEK), Tsukuba, Ibaraki 305, Japan b CTU-Computer Center, Zikova 4, 16635 Prague 6, Czech Republic
Ten years have passed since the current control system started the operation of the TRISTAN accelerator. The system uses CAMAC as a front-end electronics, and they are controlled by 25 Hitachi process computers linked by a N to N token ring network. In order to have the ability to perform complicated accelerator operations, there is a strong request to renew these 25 process computers. Firstly, we review how we will rejuvenate the current control system under some constraints, such as the lack of man-power, limited time and financing. This is followed by proposals for the next step of rejuvenation.
1. Introduction The problems of the current main control system were discussed at I C A L E P C S ' 9 1 [1]. In addition to the difficulty of maintenance, there is an urgent requirement to use the S A D (Strategic Accelerator Design) [2], which is the set of programs for accelerator simulation and optics matching calculations for operation. In particular, the study of K E K future projects, such as K E K B-factory [3] and T R I S T A N Super Light Facility [4] shows the need for very complicated calculations. In order to satisfy these requirements, the rejuvenation of the system started by rebuilding the operational consoles and preparing the interface to the SAD.
2. The rejuvenation status of the TRISTAN accelerator control system 2.1. Overview Because of the recent drastic improvement in the calculation speed of workstations, it is now possible to obtain more C P U , memories, disks, etc with lower cost and reasonable reliability. Some workstations have been added to the T R I S T A N control system. As shown in Fig. 1, the current control system consists of three systems; i.e. the main control system, the R F control system and the second control system. The main control system is composed of 25 Hitachi process computer, model HIDIC'80s. The R F control system is composed of 7 V A X stations which are used for the aging of R F cavities. The second control system is
composed of Unix workstations, V M E systems and X terminals. These control systems are connected to the Linac control system, the main frame computer (Hitach H I T A C , Fujitsu F A C O M ) in the K E K computer center and the workstation cluster for accelerator simulation (SAD). F D D I network which were laid along the T R I S T A N tunnel also connect to the second control system. 2.2. The workstation cluster for accelerator simulation
study The S A D codes have been developed in the K E K accelerator department, and are used for the design of many accelerators. They were developed on the main frame and, last April, all codes were moved to the Unix workstation cluster. Although we need large C P U power, the size of the data we treat is relatively small. So, the decision was taken to move them to the high speed workstation cluster. Since the S A D codes are written in Fortran, the installation has been smooth. The Unix workstation cluster is composed of 4 Hewlett-Packard workstations, 9000/755 and 9 0 0 0 / 735, and they are connected by an F D D I network to each other. The batch jobs are controlled by a Task broker, which is a software product of the HewlettPackard company. The interactive users are automatically distributed to the least used workstation by a simple shell program. This cluster is connected to the T R I S T A N accelerator control system via Ethernet. The kick angle of each steering magnet and the data for closed orbit distortion ( C O D ) measured by beam monitors can be stored in this system in the S A D input format.
0168-9002/94/$07.00 © 1994 - Elsevier Science B.V. All rights reserved SSDI 0 1 6 8 - 9 0 0 2 ( 9 4 ) 0 0 0 2 7 - 5
129
T. Mimashi et aL /Nucl, Instr. and Meth. in Phys. Res. A 352 (1994) 128-130 2.3. The T R I S T A N second control system
As discussed at the ICALEPCS'91, the critical problems occurred in the operational consoles. To solve these problems, we started the system rejuvenation at this level. Unix workstations and X terminals were chosen as consol hardware. Programs are written in C. Motif and PVWAVE, which is a commercial product of Highland Software, Inc. are used as the graphics software. Motif is used to build the operational consol and PVWAVE is used to plot some data, such as measured orbit position and the kick angle of each steering magnet. The advantage of the graphics based on X window is that it fits a multi-computer system. One console is composed of one X terminal and a workstation. 2.4. The communication between T R I S T A N second control system a n d the main control system
Since HIDIC'80s only support their propriety network, the communication path between the new work-
station-based system and the old Hitachi H I D I C 80 system has been made using a C A M A C memory module as a mail box. Three sets of programs are running on three different types of machine, H I D I C 80, VME and Unix workstation. A VME system is connected to the workstations by Ethernet. The program in the VME and Unix workstation is written in C and OS9 is used in the V ME system. Data is written to the CAMAC memory boards from a workstation via the VME system. The H I D I C reads the data from the CA MA C memory module, and performs such actions as reading steering data, measuring COD, setting of current of steering magnets, etc. and then returns some message to the workstation via the same path. 2.5. R F control system
The RF control system has been used only for the aging of RF cavities. This system is completely independent from the main control system. Each VAX has its own serial CA MA C driver, and each RF CAMAC crate has two crate controllers. The system will be used
TRISTAN ACCELERATOR CONTROL SYSTEM NETWORK (1993) TRISTAN Main Contro~ System (H]DIC 80)
CollgOle
TRI~AN
RF Control S y la . m (YAJ(')
C,,~pul~ c ~ l ~
I
~,.
DECS'~A'nOt~
c
° ............
V ~ A T ) O N VAXS~V6~ rt~ 4oc~o
?
VA.XS~Vl~ VAXS'rAT1ON ~0o-~o ~o
i.~c~ov.:oo~
U~
/ ]~Ibor:llor7 N d w o r k TRISTAN - LINAC
Network
Fig. 1. Present TRISTAN accelerator control system.
II. STATUS & SYSTEMS ARCHITECTURE
130
T. Mimashi et al. /Nucl. Instr. and Meth. in Phys. Res. A 352 (1994) 128-130
not only for the aging of RF cavities, but also for actual accelerator operation.
be changed from the TRI STA N control room. Two D E C workstations are used for this system and programs were developed using Motif and C.
2.6. Operational software The operational software is data driven. A set of action has been defined in 8 steps; get some information from the operator, read the necessary data, send some information to the HIDIC, receive the reply from H ID IC system, perform some calculation on the data, write the output data, run SAD if necessary, and finally, the display result. The application programer only has to fill in the necessary information. In the system, the graphics, some calculations, data storage, etc., are all taken care of on the Unix workstation. The process computer HIDIC'80 only takes care of the communications to the accelerator equipment through CAMAC, such as s e t / r e a d steering magnet currents, measure beam orbit, etc. This is the first step of the computer replacement. Firstly we try to minimize dependence on the old computers and secondly to rewrite a minimum set of the old computer functions.
2. 7. High speed data logging system Using workstations and the V M E system, high speed data taking becomes possible. For example, using a single turn beam monitor, the beam information for each revolution can be acquired and stored. Some single turn beam monitors are installed. Although some of them are still under development, it is possible to obtain transverse and longitudinal information from each bunch on each revolution. Data from these one turn beam monitors will be sent, via FDDI, to the main control room and analyzed.
3. The rejuvenation plan of the control system The next step of rejuvenation will be to move other functions performed by the H I D I C to the workstation system step by step. The study of the protocol between the workstations and the H I D I C is important. Since it becomes possible to take turn-by-turn data, it is necessary to think how to treat such a large amount of data with a data base. Such research and development must be not only useful for the TRISTAN accelerator operation, but it must also take a very important role for the next accelerator project.
4. Conclusion The rejuvenation of the T R I S T A N accelerator control system has been carried out during the last two years. As discussed in the previous ICALEPCS conference, we used standard hardware, and bought a vendor-independent software package so that laboratory staff could concentrate on writing application software. The Unix workstations X terminals and VME systems are used as new hardware components. Motif and PVWAVE which is a commercial software based on the X Window system are used for graphics and the standard languages, C and F O R T R A N , are used for programing. The speed for graphics has improved considerably. SAD comes to be used for the TRI STA N accelerator operation more closely.
2.8. The communication between T R I S T A N control system and Linac control system
References
The communication between Linac control system and T R I S T A N control system has been established. Injection beam energy from the linear accelerator can
[1] [2] [3] [4]
T. Mimashi et al., KEK Proc. 92-15 (1992) 85. K. Oide, Nucl. Instr. and Meth. A 276 (1989) 427. S. Kurokawa et al., KEK Report 90-24. S. Kamada et al., KEK Progress Report 92-1.
NUCLEAR INSTRUMENTS & METHODS IN PHYSICS RESEARCH Sect,onA
The rejuvenation status of TRISTAN accelerator control system T. M i m a s h i a, A. A k i y a m a a, S. Araki a, K. K u d o h ~, I. K o m a d a ~, T. K a w a m o t o a, S. K u r o k a w a ", T. N a i t o h a, S. T a k e d a a, j. U r a k a w a ~, T. T a k a s h i m a a, K. F u r u k a w a ~, J. Navratil b K. O i d e a N. Y a m a m o t o a a National Laboratory for High Energy Physics (KEK), Tsukuba, Ibaraki 305, Japan b CTU-Computer Center, Zikova 4, 16635 Prague 6, Czech Republic
Ten years have passed since the current control system started the operation of the TRISTAN accelerator. The system uses CAMAC as a front-end electronics, and they are controlled by 25 Hitachi process computers linked by a N to N token ring network. In order to have the ability to perform complicated accelerator operations, there is a strong request to renew these 25 process computers. Firstly, we review how we will rejuvenate the current control system under some constraints, such as the lack of man-power, limited time and financing. This is followed by proposals for the next step of rejuvenation.
1. Introduction The problems of the current main control system were discussed at I C A L E P C S ' 9 1 [1]. In addition to the difficulty of maintenance, there is an urgent requirement to use the S A D (Strategic Accelerator Design) [2], which is the set of programs for accelerator simulation and optics matching calculations for operation. In particular, the study of K E K future projects, such as K E K B-factory [3] and T R I S T A N Super Light Facility [4] shows the need for very complicated calculations. In order to satisfy these requirements, the rejuvenation of the system started by rebuilding the operational consoles and preparing the interface to the SAD.
2. The rejuvenation status of the TRISTAN accelerator control system 2.1. Overview Because of the recent drastic improvement in the calculation speed of workstations, it is now possible to obtain more C P U , memories, disks, etc with lower cost and reasonable reliability. Some workstations have been added to the T R I S T A N control system. As shown in Fig. 1, the current control system consists of three systems; i.e. the main control system, the R F control system and the second control system. The main control system is composed of 25 Hitachi process computer, model HIDIC'80s. The R F control system is composed of 7 V A X stations which are used for the aging of R F cavities. The second control system is
composed of Unix workstations, V M E systems and X terminals. These control systems are connected to the Linac control system, the main frame computer (Hitach H I T A C , Fujitsu F A C O M ) in the K E K computer center and the workstation cluster for accelerator simulation (SAD). F D D I network which were laid along the T R I S T A N tunnel also connect to the second control system. 2.2. The workstation cluster for accelerator simulation
study The S A D codes have been developed in the K E K accelerator department, and are used for the design of many accelerators. They were developed on the main frame and, last April, all codes were moved to the Unix workstation cluster. Although we need large C P U power, the size of the data we treat is relatively small. So, the decision was taken to move them to the high speed workstation cluster. Since the S A D codes are written in Fortran, the installation has been smooth. The Unix workstation cluster is composed of 4 Hewlett-Packard workstations, 9000/755 and 9 0 0 0 / 735, and they are connected by an F D D I network to each other. The batch jobs are controlled by a Task broker, which is a software product of the HewlettPackard company. The interactive users are automatically distributed to the least used workstation by a simple shell program. This cluster is connected to the T R I S T A N accelerator control system via Ethernet. The kick angle of each steering magnet and the data for closed orbit distortion ( C O D ) measured by beam monitors can be stored in this system in the S A D input format.
0168-9002/94/$07.00 © 1994 - Elsevier Science B.V. All rights reserved SSDI 0 1 6 8 - 9 0 0 2 ( 9 4 ) 0 0 0 2 7 - 5
129
T. Mimashi et aL /Nucl, Instr. and Meth. in Phys. Res. A 352 (1994) 128-130 2.3. The T R I S T A N second control system
As discussed at the ICALEPCS'91, the critical problems occurred in the operational consoles. To solve these problems, we started the system rejuvenation at this level. Unix workstations and X terminals were chosen as consol hardware. Programs are written in C. Motif and PVWAVE, which is a commercial product of Highland Software, Inc. are used as the graphics software. Motif is used to build the operational consol and PVWAVE is used to plot some data, such as measured orbit position and the kick angle of each steering magnet. The advantage of the graphics based on X window is that it fits a multi-computer system. One console is composed of one X terminal and a workstation. 2.4. The communication between T R I S T A N second control system a n d the main control system
Since HIDIC'80s only support their propriety network, the communication path between the new work-
station-based system and the old Hitachi H I D I C 80 system has been made using a C A M A C memory module as a mail box. Three sets of programs are running on three different types of machine, H I D I C 80, VME and Unix workstation. A VME system is connected to the workstations by Ethernet. The program in the VME and Unix workstation is written in C and OS9 is used in the V ME system. Data is written to the CAMAC memory boards from a workstation via the VME system. The H I D I C reads the data from the CA MA C memory module, and performs such actions as reading steering data, measuring COD, setting of current of steering magnets, etc. and then returns some message to the workstation via the same path. 2.5. R F control system
The RF control system has been used only for the aging of RF cavities. This system is completely independent from the main control system. Each VAX has its own serial CA MA C driver, and each RF CAMAC crate has two crate controllers. The system will be used
TRISTAN ACCELERATOR CONTROL SYSTEM NETWORK (1993) TRISTAN Main Contro~ System (H]DIC 80)
CollgOle
TRI~AN
RF Control S y la . m (YAJ(')
C,,~pul~ c ~ l ~
I
~,.
DECS'~A'nOt~
c
° ............
V ~ A T ) O N VAXS~V6~ rt~ 4oc~o
?
VA.XS~Vl~ VAXS'rAT1ON ~0o-~o ~o
i.~c~ov.:oo~
U~
/ ]~Ibor:llor7 N d w o r k TRISTAN - LINAC
Network
Fig. 1. Present TRISTAN accelerator control system.
II. STATUS & SYSTEMS ARCHITECTURE
130
T. Mimashi et al. /Nucl. Instr. and Meth. in Phys. Res. A 352 (1994) 128-130
not only for the aging of RF cavities, but also for actual accelerator operation.
be changed from the TRI STA N control room. Two D E C workstations are used for this system and programs were developed using Motif and C.
2.6. Operational software The operational software is data driven. A set of action has been defined in 8 steps; get some information from the operator, read the necessary data, send some information to the HIDIC, receive the reply from H ID IC system, perform some calculation on the data, write the output data, run SAD if necessary, and finally, the display result. The application programer only has to fill in the necessary information. In the system, the graphics, some calculations, data storage, etc., are all taken care of on the Unix workstation. The process computer HIDIC'80 only takes care of the communications to the accelerator equipment through CAMAC, such as s e t / r e a d steering magnet currents, measure beam orbit, etc. This is the first step of the computer replacement. Firstly we try to minimize dependence on the old computers and secondly to rewrite a minimum set of the old computer functions.
2. 7. High speed data logging system Using workstations and the V M E system, high speed data taking becomes possible. For example, using a single turn beam monitor, the beam information for each revolution can be acquired and stored. Some single turn beam monitors are installed. Although some of them are still under development, it is possible to obtain transverse and longitudinal information from each bunch on each revolution. Data from these one turn beam monitors will be sent, via FDDI, to the main control room and analyzed.
3. The rejuvenation plan of the control system The next step of rejuvenation will be to move other functions performed by the H I D I C to the workstation system step by step. The study of the protocol between the workstations and the H I D I C is important. Since it becomes possible to take turn-by-turn data, it is necessary to think how to treat such a large amount of data with a data base. Such research and development must be not only useful for the TRISTAN accelerator operation, but it must also take a very important role for the next accelerator project.
4. Conclusion The rejuvenation of the T R I S T A N accelerator control system has been carried out during the last two years. As discussed in the previous ICALEPCS conference, we used standard hardware, and bought a vendor-independent software package so that laboratory staff could concentrate on writing application software. The Unix workstations X terminals and VME systems are used as new hardware components. Motif and PVWAVE which is a commercial software based on the X Window system are used for graphics and the standard languages, C and F O R T R A N , are used for programing. The speed for graphics has improved considerably. SAD comes to be used for the TRI STA N accelerator operation more closely.
2.8. The communication between T R I S T A N control system and Linac control system
References
The communication between Linac control system and T R I S T A N control system has been established. Injection beam energy from the linear accelerator can
[1] [2] [3] [4]
T. Mimashi et al., KEK Proc. 92-15 (1992) 85. K. Oide, Nucl. Instr. and Meth. A 276 (1989) 427. S. Kurokawa et al., KEK Report 90-24. S. Kamada et al., KEK Progress Report 92-1.