A Research On Identifying IDC Based On Characteristics Of Traffic Behavior

IDC with its standardized room equipment,high-quality network bandwidth,a comprehensive monitoring and management of centralized one-stop service,greatly reducing the cost of small and medium enterprises information construction,and for the traditional enterprises to participate in "Internet+" to provide a convenient.These enterprises promote IDC which has the rapid development and high-speed construction.However,the construction of IDC is relatively independent,and there is no comprehensive planning and unified deployment,resulting in network resources utilization imbalance,reducing service efficiency.Therefore,there is a need for an effective way to help network operators and network planners to identify IDC in the network,thereby optimizing network resource allocation and deployment.According to this requirement,this thesis presents an IDC recognition method based on characteristic of traffic.The general idea of this thesis is to extract and select a set of characteristic parameters that can effectively distinguish the IP address of the data center from the network traffic of a large number of users.And then with machine learning technology IP address will be divided into data center IP and non-data center IP.Thirdly,build a network which consists of data center IP addresses.Finally use the community detection technology to divide the network,and then complete the identification of IDC.The results of this thesis are as follows:(1)In this thesis,through the comparison of the data center server IP traffic and non-server host IP traffic,we get six kinds of traffic characteristic attributes which reflect the difference between the server and non-server host,and a set of feature parameters including 24 flow characteristics parameters is extracted.And the classification performance test of the feature parameter set is completed.It is found that some traffic characteristic parameters have good effect on the data classification of the data center,which indicates that the traffic feature parameter set has the capabi lity of distinguishing the data center IP and the common host IP.(2)In this thesis,through the construction of data center IP connection network,using community detection technology,the problem of identify IDC into the community detection of network.By analyzing the network characteristics and functional characteristics of IDC,it is concluded that the IDC network based on stream connection has a community structure.Then,based on the idea of co-citation network,the IP connection graph of the data center is constructed according to the connection feature of flow,and the community detection algorithm is used to discover the target.Finally,this thesis uses the data center feature parameter set as input,and uses two kinds of machine learning algorithms to classify—C4.5,Naive Bayes,have achieved good results.While using two community detection algorithms—BGLL,Infomap,the validity of the method is verified.