ARCHITECTURAL AND COLLABORATIVE CONTROL SYSTEMS A COMPLEMENTARY SYMBIOSIS - Looking at todays control system one can find a wide variety of implementations. From pure architectural to collaborative control system (CCS) tool kits to home grown systems and any variation in-between. Decisions on the type of implementation should be driven by technical arguments Reality shows that financial and sociological reasons form the complete picture. Any decision has its advantages and its drawbacks. Reliability, good documentation and support are arguments for industrial controls. Financial arguments drive decisions towards collaborative tools. Keeping the hands on the source code and being able to solve problems on your own and faster than industry are the argument for home grown solutions or open source solutions. The experience of many years of operations shows that which solution is the primary one does not matter, there are always areas where at least part of the other implementations exist. As a result heterogeneous systems have to be maintained. The support for different protocols is essential. This paper describes our experience with industrial control systems, PLC controlled turn key systems, the CCS tool kit EPICS and the operability between all of them.- INTRODUCTIONProcess controls in general started at DESY in the early 80th with the installation of the cryogenic control system for the accelerator HERA (Hadron- Elektron-Ring-Anlage). A new technology was necessary because the existing hardware was not capable to handle standard process controls signals like 4 to 20mA input and output signals and the software was not designed to run PID control loops at a stable repetition rate of 0.1 seconds. In addition sequence programs were necessary to implement startup and shutdown procedures for the complex cryogenic processes like cold boxes and compete compressor streets. Soon it was necessary to add interfaces to field buses and to add computing power to cryogenic controls. Since the installed D/3 system1 only provided an documented serial connection on a multibus board, the decision was made to implement a DMA connection to VME and to emulate the multibus boards functionality. The necessary computing power for temperature conversions came from a Motorola MVME 167 CPU and the field bus adapter to the in house SEDAC field bus was running on an additional MVME 162. The operating system was VxWorks and the application was the EPICS toolkit. Since this implementation was successful it was also implemented for the utility controls which were looking for a generic solution to supervise their distributed PLCs.A SELECTION OF PROCESS CONTROL SYSTEMS AT DESYDCS (D/3) As a result of a market survey the D/3 system from GSE was selected for the HERA cryogenic plant. The decision was fortunate because of the DCS character of the D/3. The possibility to expand the system on the display- and on the I/O side helped to solve the increasing control demands for HERA. The limiting factor for the size of the system is not the total number of I/O but the traffic on the communication network. This traffic is determined by the total amount of archived data not by the data configured in the alarm system. The technical background of this limitation is the fact that archived data are polled from the display servers whereas the alarms are pushed to configured destinations like alarm-files, (printer) queues or displays. SCADA Systems with DCS Features (Cube) The fact that the D/3 system mentioned above had some hard coded limitations with respect to the Y2K problem was forcing us to look for an upgrade or a replacement of the existing system. As a result of a call for tender the company Orsi with their product Cube came into play 2. The project included a complete replacement of the installed functionality. This included the D/3 as well as the integration of the DESY field bus SEDAC and the temperature conversion in VME. The project started promising. But soon technical and organizational problems were pushing the schedule to its limits which were determined by the HERA shutdown scheduled at that time. The final acceptance test at the vendors site showed dramatic performance problems. Two factors could be identified as the cause of these problems. The first one was related to the under estimated CPU load of the 6th grade polynomial temperature conversion running at 1 Hz. The second one was the additional CPU load caused by the complex functionality of the existing D/3 system. Here it was underestimated that each digital and analog input and output channel had its own alarm limits in the D/3 system. In a SCADA like system as Cube the base functionality of a channel is to read the value and make it available to the system. Any additional functionality must be added. Last not least the load on the network for polling all the alarm limits typically for a SCADA system was also driving