Dependent Task List Scheduling Algorithm with Function Configuration in Edge Computing

In Mobile Edge omputing (MEC), the applic-ations offloaded from mobile devices to edge servers become more complicated, consisting of multiple interdependent tasks, and which are typically modeled as a Directed Acyclic Graph (DAG). But due to the limited capacity of resources, all functions cannot be configured in one server, so the usual practice is to configure functions on-demand when tasks arrive. We refer to the idea of classic heuristic list scheduling algorithm and jointly consider the problem of dependent task scheduling in MEC and the function configuration on edge servers. Our objective is to minimize the application completion time. Specifically, before the tasks start scheduling, we perform virtual function configuration work to determine the candidate execution edge servers for each task. Then, like the list scheduling algorithm, we define the priority calculation method, which jointly considers the computing overhead and communication overhead of the task. Finally, we propose a new heuristic algorithm, named DDoc, to determine the scheduling position and execution time of the task, and propose the function configuration strategy. Our extensive experiment on the DAG synthesis graphs show that the DDoc algorithm effectively reduces the completion time of the application. Moreover, DDoc consistently performs well on various setting of key parameters.