Structuring an EPICS System to Optimize Reliability, Performance and Cost

Conference · Mon Oct 10 04:00:00 EDT 2005

OSTI ID:850523

Bickley, Matthew; White, Karen S

Thomas Jefferson National Accelerator Facility (Jefferson Lab) uses EPICS as the basis for its control system, effectively operating a number of plants at the laboratory, including the Continuous Electron Beam Accelerator (CEBA), a Free Electron Laser (FEL), Central Helium Liquefier, and several ancillary systems. We now use over 200 distributed computers running over a complex segmented network with 350,000 EPICS records and 50,000 control points to support operation of two machines for three experimental halls, along with the supporting infrastructure. During the 10 years that EPICS has been in use we have made a number of design and implementation choices in the interest of optimizing control system reliability, performance, security and cost. At the highest level, the control system is divided into a number of distinct segments, each controlling a separate operational plant. This supports operational independence, and therefore reliability, and provides a more flexible environment for maintenance and support. The control system is relatively open, allowing any of the 300 account holders to look at data from any segment. However security and operational needs mandate restricted write access to the various control points based on each user's job function and the operational mode of the facility. Additionally, the large number of simultaneous users, coupled with the amount of available data, necessitates the use of throttling mechanisms such as a Nameserver, which effectively reduces broadcast traffic and improves application initialization performance. Our segmented approach provides natural boundaries for managing data flow and restricting access, using tools such as the EPICS Gateway and Channel Access Security. This architecture enables cost optimizations by allowing expensive resources, such as Oracle, to be effectively shared throughout the control system, while minimizing the impact of a failure in any single area. This paper discusses the various design decisions, and how Jefferson Lab has chosen to balance the often competing interests associated with operating and supporting the control system for a large machine.

• Research Organization: TJNAF (Thomas Jefferson National Accelerator Facility, Newport News, VA)
• Sponsoring Organization: USDOE
• DOE Contract Number: AC05-84ER40150
• OSTI ID: 850523
• Report Number(s): JLAB-ACO-05-431; DOE/ER/40150-3609
• Country of Publication: United States
• Language: English