基于多核处理器的边缘数据采集分析节点架构  

Edge-Oriented Data Acquisition and Analysis Node Architecture Based on Multi-Core Network Processor

作　　者：张润滋[1,2] 王劲林 陈晓[1,2] 叶晓舟 ZHANG Runzi;WANG Jinlin;CHEN Xiao;YE Xiaozhou(National Network New Media Engineering Research Center,Institute of Acoustics,Chinese Academy of Sciences,Beijing,100190,China;University of Chinese Academy of Sciences,Beijing,100049,China)

机构地区：[1]中国科学院声学研究所国家网络新媒体工程技术研究中心,北京100190 [2]中国科学院大学,北京100049

出　　处：《网络新媒体技术》2020年第2期23-28,共6页Network New Media Technology

基　　金：中国科学院声学研究所创新项目:水下大数据分析技术探索项目(编号:Y655011432)。

摘　　要：随着大数据及网络处理技术的日臻完善,云计算已经成为互联网时代软硬件部署的首选架构。同时,针对有低延迟要求的流式数据处理场景以及受限的网络带宽资源现状,边缘计算模型已成为云计算模型的有益补充。本文提出一种基于多核网络处理器的边缘数据采集分析节点架构,通过在多核系统上部署异构操作系统,实现高效的数据采集与灵活的数据分析算法的部署。分别设计了数据采集单元、数据分析单元,并通过多核并行设计、基于亲和性的数据包调度、流式封装结构及聚合日志发送等优化策略,提升了系统的处理性能。实验结果表明,所提边缘数据采集分析节点架构能够满足高速网络数据实时采集与分析的性能需求。With the tremendous evolution of relevant technologies,cloud computing has come into the first choice when deploying Internet applications.Meanwhile,in the scenario with the need of streaming processing and restricted network bandwidth,edge computing model supplements cloud computing in certain ways.We present an edge-oriented data acquisition and analysis node architecture based on multi-core network processor.A heterogeneous operating system is deployed on a multi-core system to achieve efficient data acquisition and flexible data analysis algorithm deployment.The realization structures of data acquisition unit and data analysis unit are designed respectively.On the other hand,through optimizing strategies of multi-core parallel design,flow scheduling based on affinity,streaming encapsulation structure and aggregation log delivery,the processing performance of the system is improved.The experimental results show that the proposed software architecture can meet the demand of high-speed network data acquisition and analysis.

关 键 词：边缘计算 网络数据挖掘 多核网络处理器 数据采集与分析 

分 类 号：TP332[自动化与计算机技术—计算机系统结构] TP311.13[自动化与计算机技术—计算机科学与技术]