Acquisition And Analysis Of Network-link Traffic Base On SFlow

With the enlargement of network scale and increase of net-services, the network system is becoming ever more complicated in the large and medium-sized enterprises and telecom operators. Therefore, to divide network resources from various angles becomes necessary and pressing for net management. Every network resource object should be supervised on time from multi-level and multidimensional aspects so as to provide elaborate operation information of the whole network, by which network monitors could easily find the blocking factors of low network efficiency, and then greatly improve the applicability and efficiency of network.Although many enterprises enhance themselves by network systems of multi-function, network operation administration and maintenance still lies in the awkwardly passive and uncontrollable situation. The management of network links can only reflects topology network state and some basic state of key-links (such as whether it is working, utilization rate of port bandwidth). Network monitors cannot get knowledge of the actual working specifics and quality at all, which could lead to low efficiency and unstable function. The unknown details and failures of certain level links could also be caused by the problem mentioned above.This thesis divides network system from the angle of network links. It discusses how to collect data information, how to gather as well as how to analyze flow data of network links. It also deals with how to supply a real-time performance and fore past performance report of network links. It could supply network monitors the actual situations of network links, so as to help them find the solution of network inefficiency, instability as well as some other network failures. This thesis, according to the practical demands of network performance management system developed by some company located in Wuhan, designs and accomplishes the accurate collection and analysis of network link flow. The innovative ideas and main contents of this thesis going as follows:1, systematically analyze SNMP, RMON, NetFlow network management protocol, specifically present the data collection method of sFlow, efficiently store mass data to Berkeley DB in assistant of multi-procedure technology, successfully achieve the data collection and storage structure based on sFlow2, according to the practical demands of mass data collection, analyze the deficiency of binary order tree and hashtable; successfully accomplish the collection of mass data in given time by the assistant of JudyArray and Hash table as well as quick sort arithmetic; design and achieve the real-time data mining technology of network links3, accomplish the hierarchical division of links flow information with the assistant of Struts+Spring+Hibernate structure; present the analyze result of network link flow in the form of web page...