一种基于蚁群算法的云存储副本动态选择机制研究

Pheromone-based ant colony replica selection mechanism in cloud storage

针对云存储技术中副本选择优化问题,提出一种基于蚁群原理的云存储副本动态选择算法。构建基于蚁群的副本动态选择模型,建立副本选择度量标准(如带宽占用、网络路径时延和平均访问时间等)与蚁群信息素的映射,并对虚拟机实例负载状况和虚拟机集群资源利用状况进行量化评估,感知所监控的云节点的资源度量情况;最后利用副本信息素概率计算式得到一组选择副本资源的最优解,最终达到优化负载均衡的目的。经Open Stack模式的云平台对新算法仿真实现,实验结果表明新算法成功实现了副本的有效分发和虚拟机集群的负载均衡,与Round Robin和Server Load算法相比,新算法有更好的负载均衡效果。

This paper designed and analyzed a feasible,distributed,ant colony optimization algorithm based replica selection method on big data transferring in cloud storage, which was called pheromone-based ant colony replica selection algorithnl in cloud storage （PARSA）. PARSA provided ＇a new, promising direction in cloud storage. This algorithm was different from the previously studied researches in three ways to make it more accurately reflective of cloud environment. It inclined multifactors to affect the efficiency of data accessing and processing in cloud, considered such as bandwidth, file accessing time and average delay as pheromones in PARSA. PARSA avoided a large-scale fiat flooding and supports multi-attribute range query, which were accompanied by unstructured or structured P2P replica selection methods. Lastly, simulation results from cloud test bed based on OpenStack show that, compared by Round Robin and Server Load,PARSA can reduce data access latency and bandwidth consumption, and effectively achieve cloud load balancing between storage nodes and improve the speed of data access.