Efficient Extraction Technology Of Network Flow Relationship

With the rapid development of Internet technology,the scale of China’s Internet,Internet users and the network bandwidth is increasing,various data which reflects today’s Internet environment has become increasingly complex.The paper designed and implemented a network flow analysis system for receiving and processing data stream efficiently based on current network status.On the basis of the system,we carried out research network flow extraction technology and presented the data flow difference evaluation methods for the comparison of different types of network flow to identify potential differences in it.We design and implement a user group common interest correlation algorithm based on the study of network flow extraction technology research relationship.This algoritlun can achieve the purpose of mining user additional nodes in the network flow characteristics of interest based on the associated node.The main contents of the article can be summarized as follows:1.Efficient real-time data stream processing method.The paper desigyned and implemented a universal acceptance network flow analysis framework for Communications logs,Traditional socket API features are simple and easy to maintain,while optimizing the use of multiple network adapters,no critical security buffer design and load distribution technology,we achieved the efficient flow of data receiving and processing.After the experimental tests,the system can reach to 460 MB/S of receiving rate.2.Data flow difference evaluation method.This method evaluates different types of network service flow from two angles sessions and nodes.It mainly involves the number of sessions between the primary and key node in the unit interval,uplink and downlink byte unit time between the primary and key node spacing generated,between the primary node and other characteristics associated with long connecting node to communicate the average single unit time interval.Studies have shown that the characteristics of these associated nodes can effectively reflect the communication between the primary node characteristics associated with the node,this method can identify and extract the difference between network flows.3.User-related research and implement algorithms.Design and implement common interest user groups associated algorithms to "group"in the form of network traffic generated by the user to analyze.The theory behind this algorithm is that:when a user group associated with the presence of multiple high intersection nodes in a certain time period,there is some kind of relationship between certain nodes.Mining specified user groups through the user group of public interest association algorithm node to find public interest,it helps the identification of unknown network services.Experiments show that the algorithm can effectively mining public interest of user groups node and proven these nodes have certain correlation.