Mapreduce Optimization Of Network Traffic Diversion Dedign And Implementation

The network information in the 21 st century is booming with its own unique steps.However,due to the rapid development of network technology update,Internet resources are in exponential growth rate over the information to the corners of the world.However,resources will inevitably lead to the proliferation of network congestion.Many researchers proposed the joint use of different processing nodes together constitute a distributed architecture environment to solve the load problem nodes.As traffic flow shunt in which a critical step classification.However the split is now time-consuming technical bottleneck.Cloud computing not only has a simple and transparent development model,in resolving network shunt also provides a more convenient way.In which the company proposed by the Apache Hadoop Big Data platform is a typical open source cloud computing platform.Hadoop platform mainly covers two systems,one Mapreduce programming model one HDFS distributed file system.Both work together to complete the data processing and analysis.However,the level of efficiency of data processing relating to the assessment of the performance data processing platform.Thereby optimizing the performance of the data distribution is proposed to solve the data processing efficiency is a major breakthrough.In this paper,Mapreduce simple programming model can achieve parallel computin.Meanwhile HDFS distributed file system provides a storage block,in order to improve the capacity of data processing and analysis.In this paper,the research content mainly focuses on the aspects as following: Firstly,it uses high speed data catching tool PF_RING to collect data from network as the data source,and stores it distributedly.Secondly it use the tools of the net security developing to restructure TCP/IP pckages,and restore the HTTP application layer.And then analysing the occupancy of applications by calling Hadoop clusters,moreover,it uses the distributed Mapreduce programs to analyze the occupancy.Finally realize the algorithm in the POX platform of the openflow controller.Using the form of random low in the modified Minnet test platform in order to analysis link rate,band with utilization,average packet delay of flow to verify the second proposed algorithm is better than the first algorithm in shunt aspects.