Fault-Tolerating Edge Computing with Server Redundancy Based on a Variant of Group Degree Centrality

Abstract

In the distributed and dynamic edge computing environment, edge servers are subject to runtime failures. Therefore, edge servers in an area must be fault-tolerated to ensure the reliability of services deployed on those edge servers. Server redundancy is an effective fault tolerance technique and has been widely applied in different distributed computing environments in the past decade. However, conventional fault tolerance techniques are not suitable for edge computing which has unique characteristics, i.e., the constrained coverage areas of individual edge servers (coverage constraint) and the partial overlapping between edge servers’ coverage areas (overlapping constraint). In this paper, we make the first attempt to investigate and tackle the novel edge server redundancy (ESR) problem. We prove that the ESR problem is \(\mathcal {NP}\)-hard. Then, we introduce a novel optimal approach for identifying a group of edge servers to be redundant. The objective is to maximize the effectiveness of fault tolerance measured by the harmonic mean of the scope and strength of fault tolerance given a redundancy budget. Furthermore, we propose a heuristic approach for finding sub-optimal fault tolerance strategies efficiently in large-scale ESR scenarios. Extensive experiments are conducted on a widely-used real-world dataset to evaluate the proposed approaches against three representative baseline approaches.