基于组合定权法的负载评价模型研究

Investigation on Load Evaluation Model with Combination Weighting Method

随着互联网技术的快速发展以及海量数据的几何级增长,分布式存储等相关技术得到广泛应用。其中分布式文件系统的应用为海量文件存储和管理等提供了解决方案。当前主流的分布式文件系统,如Ceph、GlusterFS、MogileFS等因为各自的特点而得到广泛应用。针对评价开源分布式文件系统FastDFS负载参数指标单一,无法实时准确反映系统中各服务器实际工作负载状态这一问题,在研究现有负载均衡技术的基础上,提出了一种基于多指标的负载评价模型。该模型既考虑了主观因素,又考虑了客观因素,运用组合定权法确定各指标权重,使得模型的权重系数能够准确反映FastDFS在不同应用场景下各服务器的真实负载。仿真实验结果表明,所提出的评价模型是有效可行的,既实现了负载的合理分布,也提高了系统的资源利用率。

With the rapid development of Intemet technology and the geometric growth of massive data, the distributed storage and other related technologies are rapidly developed. The application of distributed file system has been a solutions for massive file storage and man- agement. The current mainstream distributed file systems, such as Ceph, GlusterFS, MogileFS and so on, are widely used because of their own characteristics. Owing to the problem that the index of the load evaluation for open-source distributed file system FastDFS is single and impossible to reflect the real workload status of each server in real time, according to the research on existing load balancing technolo- gy ,a multi-index based load evaluation model is proposed. It combines the subjective and objective factors and determines the weighting of each index by combination weighting method, so that the weight coefficient of the model can accurately reflect the real load of FastDFS under different application scenarios. The simulation results show that the proposed model is effective and feasible, which not only realizes the reasonable distribution of the load, but also improves the resource utilization rate of the system.