Joint Task Scheduling and Load Balancing in Computing Power Network-enabled Edge Computing Systems

Edge computing is a promising computing paradigm to enable next-generation delay-sensitive services which are not conceivable in traditional cloud-based architectures. Due to the challenges of unbalanced workloads and limited single-point resources in edge computing systems, the collaboration among different edge nodes is an important issue to realize balanced and on-demand computing capabilities of edge systems. With the continuous extension of the network, large-scale and complex networking brings new challenges to edge collaboration. Recently, Computing Power Network (CPN) has been proposed as a possible next-generation network construction to realize optimized resource allocation among multiple heterogeneous edge nodes. In this paper, a practically deployable CPN -enabled edge computing framework is proposed. We focus on realizing computing power sharing and load balancing among heterogeneous computing nodes through flexible task scheduling, which is an attempt to achieve the interworking of computing and network capabilities. Considering the dynamic and continuous changes of the system, a learning-based algorithm is proposed to give optimal scheduling decisions. Finally, the simulation results show the effectiveness and superiority of the proposed task scheduling scheme.