Optimal User Migration upon Server failures in Edge Computing Environment

Wei Du,Qiang He,Yuan Ji,Chenran Cai,Xiaoyong Zhao

DOI: https://doi.org/10.1109/ICWS53863.2021.00045

2021-01-01

Abstract:In the edge computing environment, edge servers are geographically distributed around end-users to provide highly accessible and efficient computing capacities and services. At runtime, edge servers are subject to failures for various reasons, e.g., hardware failures, system overloaded, cyber attacks. When an edge server fails, the end-users on the edge sever, i.e., edge users, must be migrated to nearby edge servers to ensure service delivery. An optimal migration strategy must maximize the number of migrated end-users while ensuring low latency for end-users. In the meantime, the service deployment cost, i.e., the cost of deploying services required by migrated end-users, must be considered. In this paper, we model the edge user migration (EUM) problem upon server failures as an Integer Programming problem, and introduce a novel, optimal approach for solving the EUM problem. We also propose a heuristic approach for finding sub-optimal solutions efficiently to large-scale scenarios. The results of experiments conducted on a real-world dataset demonstrate that our approaches significantly outperform two representative baseline approaches.