Directed Acyclic Task Graph Scheduling for Heterogeneous Computing Systems by Dynamic Critical Path Duplication Algorithm

Abstract

This paper addresses the static scheduling of a directed acyclic task graph (DAG) on a heterogeneous, bounded set of distributed processors to minimize the makespan. We first derive the lower and upper bounds on the makespan of assigning a given directed acyclic task graph on heterogeneous processors by deferent scheduling strategies. Based on the analysis, we present a new heuristic, known as Heterogeneous Dynamic Critical Path Duplication (HDCPD), for scheduling DAG on a set of heterogeneous processors. HDCPD assigns the tasks on the dynamic critical path to the suitable processors which minimize the earliest finish time for them, combining insertion-based scheduling and task duplication techniques.The comparison study by simulation on Simgrid, based on randomly generated DAG, shows that HDCPD surpasses previous approaches in terms of both quality and cost of schedules, which are mainly presented with schedule length, frequency of best result, and scheduling time metrics.

Adaptive task scheduling in IoT using reinforcement learning

“…The most important consideration to be made is to make sure that the constraint of minimizing the communication cost between the tasks scheduled on different compute hierarchies is minimized. In critical path heuristics Fei et al (2009) for minimizing the length of the critical path, the edges (representing communication costs) are removed and the tasks are clustered together (zeroing out edges) and then assigned to the single computational resource. In the case of IoT task scheduling, the tasks with large communication overhead can be clustered together and then scheduled on a powerful fog server like Jetson Tx.…”

“…In the case of IoT task scheduling, the tasks with large communication overhead can be clustered together and then scheduled on a powerful fog server like Jetson Tx. Task duplication methods (Fei et al , 2009) also can be used to reduce the communication overhead between tasks scheduled on different computational resources. Duplication will be particularly useful only when the various tasks are assigned on different compute hierarchies, i.e.…”

“…A model of such a partition scheme can be represented as a directed acyclic graph (DAG) (Fei et al , 2009), where vertices represent the tasks and the edges connecting the vertices represent the dependency between the tasks. The edges are associated with the weight, which represents the communication cost between the tasks connected by the edge.…”