Network Traffic Flow Modeling And Behavior Analysis In Complex Networks

The core value of communication network lies in the services it carried due to the significance of social informatization.Flow behavior analysis focuses on flow characteristics and help us understand network regulations.Nowadays,networks are more and more complex and intelligent,at the same time,the coupling and bursty characteristics of flows are increasingly enhanced,flow behavior analysis becomes a difficult problem.By now,most studies focus on flow generations,arrivals,and distributions.On one hand,it can hardly build relations to flow transmission quality,on the other hand,these results are mostly based on a stable network environment which have limitations to some extent.Considering these deficiencies,this paper focuses on flow coupling mechanisms in complex communication networks,explores flow propagation dynamics in network bursty and failure environments,and proposes effective propagation control schemes to ensure flow reliable transmissions.The innovations and contributions of this paper are as follows:(1)A flow interaction network based on elastic mechanics is established to explore the coupling mechanisms of flow transmission qualities.To solve the QoS coupling problem,an undirected weighted flow interaction network is constructed by utilizing elastic mechanics,which is applicable to flow correlation analysis in large-scale complex networks.In this network,nodes denote flows,edges with weight denote elastic relations between flows.By simulating the structural characteristics of the network,'small-world effect' was found when flows randomly distributed and reaches a certain quantity,which indicates the close and clustered coupling relationship of flows.The flow degree distribution is uneven,there exists a small portion of flows with very high degrees,which have great influence on flows' transmission quality.(2)An influence propagation model of Internet bursty flows is established to analyze its propagation mechanisms and regularities.An influence propagation model based on coupling probability and neighbor relationship was proposed to study the propagation characteristics under different network types and situations.Simulation results show that flow bandwidth bursty strength,network congestion level,and network types will affect the propagation process,but all follows a unified rule:the influence propagation of the bursty flow is a process that greatly decreases as the neighbor order increases due to the dispersion effect in each order,and the influence on odd-order neighbors and even-order neighbors alternates positively and negatively.(3)An immune control strategy based on flow community structure is proposed to ensure reliable flow transmissions.To control the transmission bandwidth fluctuations caused by network bursty flows,an immunization strategy based on flow community structure was proposed.Most of the traditional methods control the flow bursty strength by increasing the system overhead which reduce the flow transmission efficiency.Without changing the flow transmission modes,our algorithm greatly reduces the influence scope of the bursty flow by cutting off the propagation path based on the flow community structure.Simulation result shows that 21%immunization flows can protect 97.2%flows from bandwidth fluctuations in an optimal situation.(4)Network cascade failure model based on coupled networks is established to analyze the dynamic behavior of flows.To understand the cascading failures caused by unbalanced traffic load,a cascading failure model based on flow layer/network layer coupling network was established to study the flow transmission and competition behavior.Simulation results showed that:1)the dependence of the flow on the network is proportional to the flow average shortest path length,which reflects the flow transmission efficiency;2)the network giant component can reflect the flow competition area,which under targeted attacks is much more obviously affected compared with random attacks;3)the flow competition intensity is closely related to flow degree distribution and flow congestion degree.The uneven flow distribution in the network will obviously affect the flow competition intensity in cascading failures.