Adaptive data replication in real-time reliable edge computing for internet of things

Many Internet-of-Things (IoT) applications rely on timely and reliable processing of data collected from embedded sensing devices. To achieve timely response, computing tasks are executed on IoT gateways at the edge of clouds, and for fault tolerance, the gateways perform data replication to backup gateways. In this paper, we report our study of data replication strategies and a real-time and fault-tolerant edge computing architecture for IoT applications. We first analyze how both embedded devices' storage constraints and data replication frequency may impose timing constraints on data replication tasks, and we investigate correlations between execution of data replication tasks and execution of edge computing tasks. Accordingly, we propose adaptive data replication strategies and introduce a framework for real-time reliable edge computing to meet the needed levels of data loss tolerance and timeliness. We have implemented our framework and empirically evaluated the proposed strategies with baseline approaches. We set up experiments using Industrial IoT traffic configurations that have requirements on data loss and timeliness, and our experimental results show that the proposed data replication strategies and framework can ensure needed levels of data loss tolerance, save network bandwidth consumption, while maintaining the latency performance.