A Survey of Faults and Fault-Injection Techniques in Edge Computing Systems

Abstract

Edge computing has emerged in recent years to reduce latency, conserve bandwidth, and enhance privacy for applications. As more edge computing applications are being deployed, there is a growing need to ensure fault tolerance for such systems. To achieve fault-tolerant edge computing, understanding potential faults and fault-injection methods is crucial. In this paper, we surveyed 76 research publications in over 25 top conferences and journals published during the last 6 years with a focus on fault-tolerant edge computing. Through our analysis, we identified the most relevant faults as well as the common practical fault-injection techniques, simulation frameworks, and benchmarks used in edge computing. Our key insights from this survey underscore the importance of handling edge-centric faults including performance, resource-stressing, and network-partition faults as well as developing edge-centric fault-injection methods and dependability benchmarks in future studies.