Knowledge reuse in edge computing environments

To cope with the challenge of managing numerous computing devices, humongous data volumes and models in Internet-of-Things environments, Edge Computing (EC) has emerged to serve latency-sensitive compute-intensive applications. Although EC paradigm significantly eliminates latency for predictive analytics tasks by deploying computation on edge nodes’ vicinity, large scale infrastructure still huge inescapable burdens required resources. This paper introduces a novel where nodes effectively reuse local completed computations (e.g., trained models) at network edge, coined as knowledge reuse. Such releases burden from individual nodes, they can save resources relying reusing various regression classification). We study feasibility our involving pair-wise (dis)similarity metrics among datasets over based statistical learning techniques (kernel-based Maximum Mean Discrepancy eigenspace Cosine Dissimilarity). Our is enhanced computationally lightweight monitoring mechanisms, which rely Holt-Winters forecast future violations updates reused models. mechanisms predict when ‘borrowed’ are insufficient being reused, triggering new process finding more appropriate be edge. provide comprehensive performance evaluation comparative assessment algorithms different experimental scenarios using real synthetic datasets. findings showcase ability robustness maintain up-to-date trading off quality resource utilization.