基于数据中心流量特征的端到端流量估计算法

发布时间：2018-05-01 13:41

本文选题：数据中心网络 + 网络测量　；参考：《计算机科学》2017年02期

【摘要】：数据中心是云计算等大型分布式计算服务的基础,有效地设计与管理数据中心需要遵循数据中心网络的端到端流量特征。然而直接地测量网络的端到端流量需要耗费巨大的软件成本和硬件成本,并且由于数据中心网络结构的特殊性,传统的计算机网络采用的流量估计方法也无法适用于现有的数据中心网络。为解决以上问题,首先依据数据中心的资源分配和链路利用率情况提取出网络的粗粒度流量特征,在此基础上提出一种基于重力模型和网络层析技术的数据中心端到端流量估计算法。与现有的流量推理算法Tomogravity和ELIA在NS3搭建的不同规模的数据中心网络中进行性能对比,实验结果表明,所提算法能有效地利用提取出的粗粒度流量特征,在保证计算效率的前提下将计算准确度大幅提升,可满足当前数据中心网络实时获取端到端流量数据的需求。
[Abstract]:Data center is the foundation of large-scale distributed computing services such as cloud computing. To effectively design and manage data center, it is necessary to follow the end-to-end traffic characteristics of data center network. However, directly measuring the end-to-end traffic of the network requires huge software and hardware costs, and because of the particularity of the network structure of the data center, The traffic estimation method used in traditional computer networks is also not suitable for the existing data center networks. In order to solve the above problems, firstly, the coarse-grained traffic characteristics of the network are extracted according to the resource allocation and link utilization of the data center. On this basis, an end-to-end flow estimation algorithm based on gravity model and network chromatography is proposed. Compared with the existing traffic reasoning algorithms Tomogravity and ELIA in the different scale data center network built by NS3, the experimental results show that the proposed algorithm can effectively utilize the coarse-grained traffic features extracted by the proposed algorithm. Under the premise of ensuring the computing efficiency, the accuracy of the calculation can be greatly improved, which can meet the demand of the current data center network to obtain end-to-end traffic data in real time.
【作者单位】：安徽农业大学信息与计算机学院;
【基金】：国家自然科学基金(61402013,61203217) 安徽省教育厅自然科学基金资助项目(KJ2014A074) 安徽省自然科学基金(1608085QF126) 江苏省无线传感网高技术重点实验室开放课题(WSNLBKF201506)资助
【分类号】：TP393.06;TP308

【参考文献】