Making Big Data Intelligent Storable at the Edge: Storage Resource Intelligent Orchestration
Making Big Data Intelligent Storable at the Edge: Storage Resource Intelligent Orchestration
复制标题
DOI：
10.1109/globecom38437.2019.9013942
复制
发表时间：
2019-12
期刊：
影响因子：
--
通讯作者：
Fuli Qiao;M. Dong;K. Ota;Siyi Liao;Jun Wu-;Jianhua Li
中科院分区：
文献类型：
--
作者：
Fuli Qiao;M. Dong;K. Ota;Siyi Liao;Jun Wu-;Jianhua Li
网络边缘设备产生了大量快速增长的数据，给异构网络的协作带来了沉重的负担。由于边缘计算应用场景的多样性，对统一的数据存储管理提出了许多新的要求，如延迟和处理效率。传统的集中式云存储在数据量激增的情况下，已经无法满足边缘计算的按需。因此，需要一个统一的存储体系结构的计算卸载方案和存储优化算法的当前改进。为了解决这些挑战，实现数据的智能化协同存储，提出了一种新的边缘云大数据统一存储架构，支持边缘服务，实现Hadoop的边缘扩展。提出了边缘节点的功能，通过基于流行度的Q-学习，同步同一邻域的边缘节点，动态存储数据，以减轻网络负载压力，提高边缘服务的效率。提出了一种通过数据边缘存储影响服务质量的智能方案，以改善资源调度和存储空间分配。仿真结果表明，该智能体系结构的优点和效率是优于比较方案的上级。
Network edge equipment has generated a large amount of fast- growing data, which has placed a heavy burden on the collaboration of heterogeneous networks. Due to the diversity of edge computing application scenarios, many new requirements are advocated for unified data storage management, such as latency and processing efficiency. Traditional centralized cloud storage can no longer meet the on- demand of edge computing in the case of a surge in data volume. Therefore, a unified storage architecture is required for the current improvements in computational offloading schemes and storage optimization algorithms. To solve these challenges and make data intelligent collaborative storable, this paper proposes a novel unified storage architecture for big data in the edge-cloud, which supports edge services in order to extend Hadoop at the edge. The functions of the edge nodes are proposed to synchronize the edge nodes of the same neighborhood and store data dynamically via Q- learning based on popularity, in order to mitigate network load pressure and improve the efficiency of edge services. An intelligent scheme that impacts the quality of service (QoS) through data marginal storage is proposed to improve the resource scheduling and to the distribution of storage space. Simulation results demonstrate the merits and efficiency of the proposed intelligent architecture is superior to the comparison schemes.