一种基于云计算的网络流量分析系统结构

A network traffic analysis system architecture based on cloud computing

基于云计算平台Hadoop提出一种新的分布式网络流量分析系统结构。在关键监测点附近部署小型本地云,采集的流量发送到本地云进行存储和分析,本地云规模可根据监测点流量大小按需配置;分析结果传输到协调节点,存入关系数据库供查询;协调节点调度分析任务在各本地云上运行。新结构具有统一的并行处理编程框架,且能减小大量数据传输对被监测网络性能的影响。在小型云平台上用实际流量数据验证用Map-Reduce程序对分组进行统计分析的性能,相对于顺序程序处理,执行速度可提高90%以上,表明用小型云实现快速海量流量分析的方案是有效的。

Based on the cloud computing platform Hadoop,a novel architecture for distributed network traffic analysis system is proposed.A small size cloud,called local cloud,is deployed near the selected pivotal network device,and traffic collected from the device are stored and analyzed in the cloud.The size of the cloud could adapt with the traffic volume of the monitored device.The analysis results are transferred to a coordinating node,which stores them into a relation database for querying.The coordinating node schedules analysis tasks to execute in local clouds.A system in this architecture has a unified parallel programming framework,and could alleviate the impacts of large scale data transmission on performance of the monitored network.Real packet traces are used to verify the performance of statistic analysis on small size cloud.The results show that comparing against sequential analysis program,the performance of Map-Reduce program is improved by more than 90%.Therefore,it is effective to analyze larger scale network traffic using a small size cloud.