Six collaborating laboratories are building the SNS at Oak Ridge. They are addressing the issues of organization, communication, standardization and integration that are inherent to such a widely distributed project. LIGO requires many multiple input multiple output (MIMO) control loops. Its control system integrates project-designed data acquisition, commercial controls and EPICS. The CERN experiments ATLAS and CMS are building control systems based on the commercial Supervisory Control and Data Acquisition (SCADA) system PVSS II. CMS reported on integrating Oracle in a SCADA to store the large amount of calibration data and front-end electronic configuration data needed for the detector. ALMA will consist of 64 12 m submillimetre antennas, with baselines up to 10 km, located at an altitude of 5000 m in the Chilean Atacama desert. Its control system incorporates JAVA and CORBA.
Future trends
The trend towards ever more geographically distributed collaborations means that development and operation teams are also increasingly widespread. Here, the SNS development team is providing a useful case study. Although feasible, development in this way requires much travel on the part of the coordination team and efficient project management to resolve potentially competing designs. Technologically, development on a global scale can be done, but the sociological and management issues are more difficult.
On the software front, the explosive growth of EPICS continues, expanding well beyond its origins in the US accelerator community. Linux seems to be the favoured operating system for servers, in particular the Redhat implementation. Ethernet is ubiquitous and is an alternative to field buses.
Adequate bandwidth is currently available and more is coming. Japan's Spring-8 synchrotron team showed interesting performance measurements of gigabit Ethernet, including its latency and redundancy. CISCO announced 10 Gbps Ethernet for 2002, with the price per Mbps dropping to $4 by 2005. Ethernets of 40 and 80 Gbps are not far off. Oracle databases, whose performance keeps improving, are used almost everywhere for storing configuration data, archived data and slow real-time data. The World Wide Web is ubiquitous and has become an integral part of the software design process. Java is increasingly being investigated as an alternative to C++. There are ample CORBA projects to demonstrate the viability of the approach. However, CORBA is complicated and there are multiple implementations and many ways to use it. As NIF engineers point out, system performance with CORBA may be a problem as it is strongly dependent on the network design.
Software engineers presently lead system architecture design, but hardware engineers are starting to enlarge the design space of system architecture. There are many smart complex device designs, which incorporate significant computing and communications capabilities. How network and system architectures will evolve when there are many such devices, each one of them a network object, remains to be seen.
Many presentations highlighted the creative use of sophisticated hardware and software tools to solve very specific and complex problems. At Argonne's Advanced Photon Source, for example, the control system incorporates accelerator orbit feedback. At Japan's KEK laboratory, beam feedback is also built into accelerator control systems. At the NIF, the control system will control wave front development; and LIGO's control system will sense and control wavelengths and frequencies.
Control systems are increasingly being taken up by industry. The SLS reported that vendors successfully delivered turnkey accelerator subsystems, including a complete linac, with EPICS-based controls as part of the contract.
The success of ICALEPCS 2001 is a tribute to the vitality of the field. The community now looks forward to the next ICALEPCS conference, which will be hosted by the Pohang University of Science and Technology in Gyeongju, South Korea, in October 2003.