SmartDis: Near-Optimal Task Scheduling in Multi-edge Networks

Abstract

In multi-edge networks, as the bandwidth and computing resources of edge servers are limited, transmission and processing of large amounts of data could bring significant pressure, leading to violations of service agreements. Thus, it is very important to schedule tasks in edge network efficiently for better performance. In this paper, we formulate the problem as minimizing the overall completion time of tasks in edge networks. Since the problem can be proved to be NP-hard, we propose a novelty algorithm, SmartDis, for scheduling tasks accross multiple edges. The main idea of SmartDis is to select offload slots of tasks based on the principle of choosing the smallest sum of added value of the overall completion time. We show theoretically that the system transmission time of SmartDis is within a constant times of the optimal result, as long as the data upload is scheduled according to the transmission order. The evaluation results illustrate that SmartDis is superior to other cross-domain job scheduling algorithms at this stage, achieving a performance improvement of at least 25%.