Design And Realization Of Distributed Network Packet Collection And Backtracking Analysis System On GPU

With the rapid development and popularization of network information technology,the network applications and traffic also presents a trend of unprecedented growth.Therefore,for huge amounts of heterogeneous data of the analysis and mining work also become extremely difficult.As the basis of network big data analysis and mining,network packets analysis can accurately predict the current network situation.This paper designs and implements distributed network data packet collection and backtracking analysis system.The system can realize the collection,storage and analysis of network packets.Distributed network data packet collection and backtracking analysis system not only can provide real network historical status data for network optimization,but also can provide basis for evaluating network status and predicting future development trend.This paper introduces the design and module development of network packet collection and backtracking analysis system.Firstly,the functional and non-functional requirements analysis of network data packet collection and backtracking analysis system are based on the related technologies of network packet collection and analysis.Secondly,according to the result of requirements analysis,the functional architecture and technical framework of the system are designed and realized,and the design and implementation of each sub-module are elaborated.Finally,the network packet collection and backtracking analysis system through the actual deployment testing,verified the system under the huge amounts of data collection,storage and analysis ability,and under the huge amounts of data query performance of the system and the system stability and extensibility.The core module of network packet collection and backtracking analysis system mainly includes the packet collection module,data packet GPU analysis module,the backtracking module and data storage module,etc.The data packet acquisition module is implemented using the Go language based on Libpcap and BPF,which can be used to filter network packets.At the same time,this system use GPU and HDFS file system and other related technologies to realize parallel parsing and storage of packets,greatly improve the performance of the network packet collection and backtracking analysis system.