K-LZF : An efficient and fair scheduling for Edge Computing servers

Abstract

With the emergence of the increasingly heterogeneous Internet of Things(IoT) devices, Edge Computing servers are required to support a variety of services with different quality of service requirements. The degree of heterogeneity of IoT devices makes it more difficult to fairly and efficiently allocate resources based on the task's weight. However, most fair schedulers are not suitable for simultaneously providing scalability and robustness in Edge Computing servers. In this paper, we propose K-LZF which is an efficient and fair scheduling algorithm for Edge Computing Servers. K-LZF aims to achieve a high level of proportional fairness for a large number of heterogeneous tasks, with constant overhead. We simulated and evaluated the performance of the proposed K-LZF in a heterogeneous IoT environment. We also designed and implemented in the AVOS kernel to show that it is applicable in actual IoT environment. The results of the simulation and implementations show that the proposed K-LZF outperforms several existing scheduling algorithm with respect to scalability and robustness even when the degree of task heterogeneity becomes high. © 2019 Elsevier B.V.