基于时间序列的 PDM 文件数据块分布算法

摘要

To meet the requirements of PDM mass data storage and high concurrent access,the PDM system based on the enter-prise private cloud and NoSQL is built,which is an effective attempt,but the default data block distribution algorithm has two major problems,nodes requests in the cluster are not even and file data block are stored centrally.Data block of PDM file distri-bution strategy based on time series analysis was presented for these problems.Using the fitting model to predict the future trend of the data block access according to the prediction results dynamically migrate data block,so achieving a reasonable distribution of data blocks.The experimental results indicate that the algorithm achieves a balanced amount of nodes access and improves fault tolerance of the cluster system.%为满足 PDM 海量数据存储与高并发访问要求，构建基于企业私有云和 NoSQL 的 PDM 系统是一个有效尝试，但现有文件系统数据块分布算法存在数据节点访问量不均衡及文件数据块存储集中两大问题，为此提出基于时间序列预测的PDM 文件数据块分布算法。根据数据块的访问量，使用拟合模型预测数据块未来的访问趋势，据此进行数据块的动态迁移，实现数据块合理分布，有效提高系统的容错性。实验结果表明，该算法实现了数据节点访问量均衡及文件数据块的均匀分布，提高了集群的容错性。