The Establishment And Application Of IPTAS Benchmark TCP Flow Data Sets

With the rapid growth of Internet scale,the behavioral characteristics of the network traffic are becoming more and more complexity.At the same time,the demand for network resources and network service quality is getting higher and higher.The current situation of network development urgently requires the managers of the network to be able to manage and control the performance of the network accurately,which can improve service quality of the network.IPTAS(iptas.edu.cn)is a network traffic analysis system developed by CERNET Southeast China(North)Regional Network Center under the support of National 973 and National Support Program.It supports network traffic collection,storage,analysis and publishing,and providing run,debug and test for all programs based on passive measurements.Since 2005,it regular collects network traffic at the boundary of CERNET in Jiangsu Province,and publicly released.Because the collection point is located at the boundary of the large-scale access network,the data saved by IPTAS is not affected by asymmetric route,and has good flow integrity.Based on the data in the IPTAS system,many scientific researches related to Internet performance can be carried out.In related research,including end-to-end transmission performance,the IPTAS system can provide an analytical data source for the algorithm,but can't evaluate the accuracy of the algorithm.To this end,the research work of this paper revolves around the establishment of IPTAS benchmark TCP flow data sets,which is used as a supplement to the data in the IPTAS system to support the verification of related scientific research.In this paper,IPTAS benchmark TCP flow data sets refers to the collection of all message sequences,which collected at the source the sink and the collection point synchronous,in the same TCP session.In order to establish the data sets,need to collect data at different geographical locations,and then,summarize,collate and match it.To this end,this paper was designed a scheme consisting of two phases.The first phase is the data collection,which focuses on collect the data at different geographical locations and summarizes them.The second phase is the data collation,which is mainly collation the data to establish the benchmark data sets.The data collection system uses the C/S structure,which consists of three parts:the client,the server and the collection point.The server uses globally unique address and specific port number.The client establishes a TCP connection with the server based on the address and port number,uploads or downloads the file and the address and port number is used as filtering rules to obtain the reference data.This paper introduces the design and implementation process of the client based on Windows system in detail.The data collation system consists of three modules:data synchronization,data collation and data standardization.The system selects the synchronization time,which based on time relationship of flow record,to synchronize the benchmark data.And then based on complete TCP flow record to match the benchmark data,and finally with the NBOS interface to complete the standardization of benchmark data.The result of the whole system shows that the above scheme can effectively establish the IPTAS benchmark TCP flow data sets.Finally,a packet loss rate algorithm(MT algorithm),which is based on IP Trace,is selected as an application example and the accuracy of the algorithm and its influencing factors are analyzed by using the standard answers constructed by the benchmark data sets.The analysis shows that the algorithm can effectively determine whether packet loss occurs,but the accuracy is bad.The analysis also shows that relative position of the measurement point and SACK will cause the probability of estimation value to be small becomes larger.