Prototype Implementation and Performance of Time-based Distributed Scheduling on Linux for Real-Time Cyber-Physical Systems

The fundamental idea of time synchronization comes from the need for executing coordinated/synchronous activities in spatially distributed systems based on time, typical applications include measurement and control systems, resource distribution, etc. The IEEE 1588 Standard for a Precision Clock Synchronization Protocol for Networked Measurement and Control Systems has been developed just for doing coordinated measurements and actuation (control). The success of this protocol is shown clearly by the wide-scale hardware and software support for it in embedded hardware and the increasing use of it in high-end Cyber-Physical Systems. However, developers must not forget that the primary aim is not time synchronization but the coordinated execution of activities (action) of applications running on the computing nodes of the distributed system, i.e., time-based distributed scheduling. Furthermore, evaluation of the properties of time synchronization is only a partial result, the performance of the application depends on the coherent execution of activities on the individual computing nodes of the distributed system. Therefore, one should evaluate how a computing platform, including hardware and software, can support coordinated activities based on synchronized time. Our paper investigates time-based distributed scheduling of activities in distributed systems that uses Linux on the nodes of the embedded system. More specifically, it evaluates how time-based distributed scheduling is supported by System on a Chip platforms and by the Linux operating system running on them, and how application developers may implement such systems using the operating system services of Linux. In addition, it presents our initial performance evaluation results on the Beaglebone Black single board computer.