A Demand-Driven Incremental Deployment Strategy for Edge Computing in IoT Network

Abstract

Edge Computing brings great opportunities to enable the Internet of Things (IoT) vision. A major challenge in this topic is the physical edge server deployment problem, which dramatically affects the service ability and service cost of edge computing. Previous work mostly assume that the edge servers are installed at one time. However, due to ever-increasing services, limited budget and evolving techniques, it is more reasonable to deploy edge servers in the IoT network in a gradual fashion. In this paper, we propose a demand-driven incremental deployment strategy (DDID) to resolve the problem. First, a novel demand model is designed to quantify the rigid and non-rigid demand of IoT services for edge computing. Then, we formulate the edge server multi-period deployment problem as a bi-level integer linear program model. The lower-level placement stage is to minimize the overall deployment cost throughout a planning horizon. We adopt a subgradient optimization with Lagrangian dual to solve this subproblem. In the upper-level allocation stage, due to the capacity limitation, we adopt a best-effort tuning scheme to prioritize the high demand services with multiple objectives. This subproblem is addressed by an improved MOEA/D algorithm. Finally, we evaluate the DDID in synthetic topologies.