Adaptive resource prefetching with spatial-temporal and topic information for educational cloud storage systems

Prefetching proactively resources at datanodes within distribution networks plays a key role in improving the efficiency of data access for e-learning, which requires assistance from semantic knowledge on educational applications and resource popularity. To capture such information, we should exploit the characteristics of education end-users’ requests with high spatial–temporal locality. This paper aims to develop resource popularity modeling techniques for enhancing the performance of educational resource prefetching. Specifically, a novel topic model, built on an accelerated spectral clustering and an ontology concept similarity, is proposed to support resource access based on semantic features, including topic relevance and spatial–temporal locality. Using the proposed model, an adaptive prefetching inference approach is presented to associate possible popular resources in the future data requests. Also, an efficient prefetching management mechanism incorporating with replica techniques is suggested to design resource cloud storage systems for geo-distributed educational applications. Experiments over a simulation setting and a real-world case study with seven million users across China are carried out. Results demonstrate that the proposed method performs favorably compared to the state-of-the-art approaches.