LET Paradigm Scheduling Algorithm Considering Parallel Processing on Clustered Many-core Processor

抄録

Embedded systems such as self-driving systems require computing platforms with high computing power and low power consumption. Multi/many-core platforms meet these requirements effectively. However, for hard real-time applications, multiple demands on shared resources can impede real-time performance, and in particular, memory is a resource that can restrict the desired performance significantly. Therefore, the logical execution time (LET) paradigm has gained attention to make the timing of memory access deterministic. This paper proposes a theoretical scheduling method for a model applying the LET paradigm to the directed acyclic graph (DAG) nodes for a multi/many-core platform. However, the LET paradigm lacks scalability owing to the overhead caused because the LET paradigm is set longer than the actual execution time of the task. The proposed method performs a parallel calculation of tasks utilizing many cores to deal with the overhead caused by adopting the LET paradigm. The evaluation shows that the proposed method benefits from the adoption of the LET paradigm and that the end-to-end latency variation is smaller than in the existing scheduling methods. Furthermore, the evaluation shows that the proposed method can maintain the task deadline miss low, despite the overhead of the LET paradigm.