Research Of Parametric Analysis And Network Status Cognitive Model In The Heterogeneous Networks

The communication networks mixed by wired networks and wirelessnetworks will become a heterogeneous collection of networks composed of large,complex, open in future. There are huge differences between the various network, sothat it is difficult to cognitive networks. The mainly different arein the underlying transmission medium and the underlying protocol in a nutshell. Theseven-layer protocol architecture design of the TCP/IP, also caused by shielding theunderlying differences, the transparent transfer of data at the same time,the network does not recognize these differences, can not do targeted traffic regulation.With the problem of TCP performance degradation in heterogeneous network,this article establish a network cognitive model of the long queue of the bottlenecklink queue, we get the estimated value of the bottleneck link queue length at thesending end. Then, Using gray relational analysis method to calculate thecorrelation of different network status parameters and the bottlenecklink queue length team, Quantitative description of the network state parameters reflectthe situation of the network performance of, Network is divided into three levels as thetarget set, select the desired rate and actual rate difference (Diff) and calm furnacequeue length is estimated that two (L) as a factor set to establish two sets ofmembership functions, the use of Fuzzy evaluation method to determine the networklevel to guide the flow control and packet loss differentiation, Join thenetwork cognitive model TCP improved version of TCP-Improved.The simulation results show that TCP-Improved protocol joined the networkcognitive model can achieve better throughput, under different background trafficloads, especially when the network load is heavy, the wireless packet loss rate, thisprotocol has better fairness and friendliness.