Selection of Service Nodes in Edge Computing Environments

Abstract

Traditional centralized mobile cloud computing is encountering severe challenges, due to the ever-increasing growth of portable smart devices, which host numerous resource-hungry applications generating exponentially growing data traffic volumes. Edge computing technology promises an increased quality of service with low communication delays by bringing the cloud closer to the end-users, at the network edge. The approach of placing server capabilities at access nodes, yielding edge service nodes, seems to be a feasible solution. In this way, proximity to service is achieved and guaranteed QoS is provided. However, keeping operational edge service nodes at every access node is energy consuming and yields increased operational expenses to the application provider when the requests for services are at low levels. In this paper, we propose algorithms to select the most appropriate access nodes to host the edge service nodes and to determine the set of access nodes that will be served by each service node, given the number of required service nodes to satisfy service computational requests and the topology of the network. We study how algorithms behave in terms of service latency and balancing of the load among service nodes, at increasing computational loads. The evaluation is performed with simulation experiments and comparison against an exhaustive search algorithm whose main goal is to minimize the latency of offered services to the access nodes. The results validate the efficiency of the proposed schemes.