Chain-Aware Scheduling for Mixed Timer-Driven and Event-Driven DAG Tasks

• Item type: Symposium(1)
• 公開日: 2022-12-20
• タイトル: Chain-Aware Scheduling for Mixed Timer-Driven and Event-Driven DAG Tasks
• タイトル / 言語: en
• タイトル: Chain-Aware Scheduling for Mixed Timer-Driven and Event-Driven DAG Tasks
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: Saitama University
• 著者所属: Saitama University
• 著者所属(英): en
• 著者所属(英): Saitama University
• 著者所属(英): en
• 著者所属(英): Saitama University
• 著者名: Daichi, Yamazaki; Takuya, Azumi
• 著者名(英): Daichi, Yamazaki; Takuya, Azumi
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Embedded systems, such as self-driving systems, periodically execute tasks to accurately recognize an external environment. To meet the deadline of embedded systems, the directed acyclic graph (DAG) is used for scheduling in existing studies. However, the DAG of the self-driving system is complex because the DAG is comprised of timer-driven tasks, triggered by period and event-driven tasks triggered by arriving input data. Because of periodic output, timer-driven tasks exist not only at the beginning of the end-to-end path but also in the middle. The existing studies have not considered the scheduling of the DAG that has multiple timer-driven tasks in the middle of the end-to-end path. To solve this problem, we propose DAG scheduling using chains. A chain is a sequence of tasks with data dependencies, and chains are triggered periodically. By dividing a DAG into chains and scheduling by chains, the proposed method can statically consider the scheduling order. Moreover, if the period of a task is larger than that of the predecessor task, the output of the predecessor task may not be used. Tasks whose output is not used are excluded from scheduling to execute other tasks. As a result of deleting tasks that do not contribute to the exit chains in the system, the schedulability is improved compared with a method based on an existing algorithm without deletion.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Embedded systems, such as self-driving systems, periodically execute tasks to accurately recognize an external environment. To meet the deadline of embedded systems, the directed acyclic graph (DAG) is used for scheduling in existing studies. However, the DAG of the self-driving system is complex because the DAG is comprised of timer-driven tasks, triggered by period and event-driven tasks triggered by arriving input data. Because of periodic output, timer-driven tasks exist not only at the beginning of the end-to-end path but also in the middle. The existing studies have not considered the scheduling of the DAG that has multiple timer-driven tasks in the middle of the end-to-end path. To solve this problem, we propose DAG scheduling using chains. A chain is a sequence of tasks with data dependencies, and chains are triggered periodically. By dividing a DAG into chains and scheduling by chains, the proposed method can statically consider the scheduling order. Moreover, if the period of a task is larger than that of the predecessor task, the output of the predecessor task may not be used. Tasks whose output is not used are excluded from scheduling to execute other tasks. As a result of deleting tasks that do not contribute to the exit chains in the system, the schedulability is improved compared with a method based on an existing algorithm without deletion.
• 書誌情報: Proceedings of Asia Pacific Conference on Robot IoT System Development and Platform; 巻 2022, p. 42-49, 発行日 2022-12-20
• 出版者 / 言語: ja
• 出版者: 情報処理学会