摘要： 针对云存储系统中数据副本管理的延时响应等问题，提出一种面向非结构化数据的分布式副本管理模型。该模型采用机架选举算法，通过提高每个机架能源利用率的方法降低系统整体能耗，为绿色数据中心提供技术保障。运用多路线性散列算法，将数据副本动态均匀地分布到不同机架的不同节点中，以提高系统性能、平衡负载和资源利用率。仿真实验结果证明，与传统的全局映射法相比，该模型可以达到较高的存储与负载平衡，具有良好的扩展性和可用性。

关键词: 分布式, 非结构化, 数据副本, 机架, 线性散, 软件事务内存

Abstract: Aiming at the problem of data replicas management time-delay in the cloud storage system, this paper puts forward a distributed replica management model for unstructured data. The model reduces the system energy consumption through carefully designed rack-selection algorithm, which is the technical basis of green data center. To improve system performance and balance utilization of resources, the model uses a multi-linear-hashing algorithm, which can balance the data replicas distribution in different nodes intelligently. Simulation experimental results show that this model can reach higher storage and load balance compared with traditional global mapping method, it has excellent system performance, scalability and availability.

Key words: distributed, unstructured, data replica, rack, linear hashing, software transactional memory

中图分类号: 

• TP311