| With the rapid development and increasing popularity of the Internet of Things(Io T)and Artificial Intelligence(AI),a vast amount of data is being generated at the network edge.Edge computing,which sinks computing and storage capacity closer to users,is a new computing paradigm with the advantages of low latency for data access and bandwidth savings in the core network.However,the limited deployment scale of edge servers compared to cloud computing results in a resource-constrained nature.Consequently,how to store and manage the vast scale of data in such an environment has become an urgent problem.Existing data storage management mechanisms are designed for cloud computing environments and are unsuitable for the characteristics of edge computing.Specifically,there are three main issues: large data storage granularity,high data redundancy due to extensive geographical deployment of edge servers,and the lack of collaboration mechanisms among edge servers.This results in load imbalances due to the heterogeneity of storage resources,ultimately increasing the cost of data storage.Furthermore,the unique attributes of the edge computing environment,including limited server coverage,low latency demand for data retrieval,regional data distribution,and heterogeneity of edge resources,make it difficult to explore cost-effective optimized data storage management mechanisms.Based on this,our research mainly focuses on achieving cost-effective optimized data storage management mechanisms in edge computing environments,minimizing data storage costs while ensuring efficient data storage management.The current research on edge data storage management mechanisms for storage cost-benefit optimization can be divided into three core strategies: data storage placement,redundant data de-duplication,and storage load balancing,with the two starting points of saving storage space before data storage and releasing storage space after data storage.In terms of how to save storage space before data storaging,the high storage overhead problem of the existing multi-copy data storage strategy is discussed,and the erasure codebased data storage strategy is introduced in the edge computing environment.The main differences between cloud storage systems and edge storage systems are analyzed,the main challenges of applying the erasure code data storage mechanism in edge storage systems are studied,and the characteristics of proximity constraint,transmission constraint,and encoding constraint of edge servers are formally defined.An optimized algorithm based on integer programming is proposed,and an approximation algorithm based on a voting strategy is developed to tackle the problem of high computational overhead.Through coding parameter optimization and placement optimization of coding blocks,storage space of the edge storage system can be saved.Experimental results on real-world datasets show that the proposed methods can save storage space by 68.58% on average compared with other comparative algorithms,and the space saving can be as high as 81.16% in large-scale scenarios.In terms of releasing storage space after data storage,this study examines the redundant data deduplication strategy for the edge computing environment.We discuss the redundancy situation of edge data and the resulting waste of storage resources,as well as the limitations of existing deduplication strategies designed for cloud storage data in the edge computing environment.Unlike the global deduplication scheme,we propose a formal definition of the edge data deduplication problem based on network topology awareness by analyzing the edge server network topology.The deduplication operation of redundant data among neighboring edge servers is performed to minimize the overall system storage cost while ensuring the coverage of data services.We propose an optimized algorithm based on integer programming and an approximate algorithm based on heuristic search to solve this problem.Experimental results on real-world datasets demonstrate that the proposed method can achieve an average data deduplication ratio of 49.87%,which significantly reduces the edge data storage cost.In terms of releasing storage space after data storage,this paper discusses the issue of data load unbalance resulting from the heterogeneity of edge resources and its impact on storage cost-effectiveness.To address this,data storage load balancing strategies are proposed for edge computing environments through collaborative resource sharing among edge servers.Redundant data is removed from some edge servers to achieve data storage load balancing.Data storage benifits are defined based on differences in data access latency to avoid aggregation of popular data.To alleviate the high computational overhead of the optimized algorithm based on integer programming,a Lagrangian relaxation-based approximation algorithm for edge data load balancing is proposed and theoretically proven to be convergent.Experimental results on real datasets show that the proposed method can achieve optimal data load balancing with an average performance improvement of 53.85%over other comparative algorithms,significantly improving storage resource utilization. |
