The multi-type DAG task scheduling method based on communication overhead

With the increasing performance of heterogeneous multi-core processors, the scheduling problem and scheduling algorithm with relational tasks become the focus of researchers. In this paper, the Directed Acyclic Graph (DAG) task model on heterogeneous multi-core platform is analyzed. Because there are many types of computing resources in heterogeneous environment and they are widely distributed, the communication overhead between tasks on different cores exists in the actual scheduling process. However, most of the current scheduling algorithms either ignore the communication overhead between tasks on different cores, or fail to consider the communication overhead of tasks between cores when deciding and selecting processor cores, resulting in waste of resources and increasing the overall length of the system scheduling tasks. Therefore, for the DAG task model with communication overhead, this paper discusses the influence of the communication overhead between tasks on the task scheduling algorithm, and proposes a Composite Node Generation Algorithm（CNGA） task scheduling algorithm. According to the amount of communication overhead between tasks, the algorithm is preprocessed and allocated to generate composite nodes, and the upstream rank and communication ratio of task nodes are used as parameters to determine the priority of task nodes. Compared with HEFT and CPOP algorithms, the experimental results show that the ratio of task communication overhead to task execution overhead Communication to Computation Ratio (CCR) under certain conditions, the response time is improved by 10.19% compared with the original HEFT and CCOP, and the stability is increased by 3.96%.