Structuring an EPICS System to Optimize Reliability, Performance and Cost

1
Structuring an EPICS System to Optimize
Reliability, Performance and Cost

• Matt Bickley and Karen White

2
Design Choices

• How available?
• How difficult to maintain?
• Is the performance sufficient?

3
First, introductions
4
Jefferson Laboratory

• Continuous Electron Beam Accelerator
• Superconducting RF LINACs
• Three-hall simultaneous delivery
• Up to 6 GeV electrons with high polarization,
  10e-4 energy spread and 100 micron spot size

5
EPICS History at JLab

• Old control system was inadequate
• Started using EPICS in 1995, in the midst of
  commissioning
• Implementation resulted from the timing
• Few front-end computers
• CAMAC-based acquisition systems

6
Optimizing Cost and Performance

• Obvious solutions to system performance
• Additional front-end computers
• Upgrading CPUs
• More RAM
• Switched network
• Nameserver

7
Nameserver Results
8
Maintenance A Scheduling Challenge

• Developing for and supporting a dynamic
  environment
• An electron factory
• Availability is the primary measure of success
• Run 35 weeks/year

9
2004 An Operational Year
10
Maintainability

• Front-end concentration vs. isolation
• Hardware costs
• Localization of control
• Resource contention
• Back-end costs
• Segmentation by operational plant
• Sharing expensive resources

11
Designing for Availability

• Tailor design to an operational schedule
• Maximize maintenance opportunities
• Optimize limited testing time
• Use accelerator as integration test bed
• Strict accountability for focusing resources
• Mean time between failure
• Mean time to repair

12
Control system MTBF and MTTR
13
Downtime Configuration, Requirements or Bug?
Hours down due to
14
Summary

• Performance choices are typically apparent
• Maintenance needs driven by operational
  scheduling
• Maximizing availability
• Careful configuration management
• Thorough, well-documented requirements