Time-triggered communication scheduling analysis for real-time multicore systems

Abstract

The demand for more computing power in current real-time systems carries on the development and research on multicore devices. Especially for hard real-time applications, like an engine control system, the software needs to be distributed and scheduled effectively. These applications consist of many tasks, which communicate data among each other. Considering a multicore system, communication between cores may require a lot of time. A bus architecture becomes a communication bottleneck with an increasing number of cores. Therefore, we consider a scalable communication structure like a Network-on-Chip (NoC). This paper studies the schedulability analysis of tangled tasks by resolving the communication dependencies with a Time-Triggered Constant Phase (TTCP) scheduler. A TTCP scheduler assigns periodic time slots for each computation and each communication entity. With the TTCP approach, we can highly utilize the NoC and the cores considering a tangled task model. However, this approach requires a method getting a feasible set of these time slots. We provide a schedulability analysis and a heuristic algorithm, that runs in pseudo-polynomial time complexity, for assigning the time slots. Experiments confirm this result and show the effectiveness of our heuristic algorithm for assigning the time slots for our approach. For typical industrial task sets with 1000 tasks, our approach can utilize the NoC by around 60%, while holding all real-time constraints.