Research On The Dissipative Structure Of Internet Macroscopic Topology

As a typical tremendous complex network, Internet calls more and more interest in characteristic analysis and modeling in academic field. In recent years, research has made considerable progress in the field, however, faced with large and complex network, not only pursuit of technical details, but also concerned about the macro-topology are essential. Only in-depth understanding of topology features and their interconnections can better construction and development of the Internet be achieved. Analysis and modeling of the dissipative structure and dynamics behavior of the Internet will help people better understand the characteristics of the Internet topology and infrastructure, which leads to great significance in the design, implementation and performance research of the next generation Internet.At first, this paper measures the CAIDA's data in the period from January 2009 to December 2011, mainly describing the recent Internet network topology characteristics by preprocessing and amending. Through the evolution of the regular network topology and the characteristic attributes of a single node, overall network topology will be understood thoroughly. The Internet consists of fewer core nodes and large part of leaf nodes. The size of the network has been increased slightly by time, the importance of the "hot" node has been reduced, the network tends to flatten, and robustness enhances. The changes in macro structure and its inherent connection mechanism can be obviously seen from the evolution of overall characteristics, in turn, mutation node will be found.Secondly, according to the definition of k-core, the internet topology was divided into different cores at AS level. Based on the cores and dynamic change, focusing on analyzing node connection distribution and dynamic change, it has been found that the number of core stabilize gradually and each shell represents power-law characteristic and obvious self-similarity. Node connection change plays an important role in network changes. Focusing on the inter-shell dynamic changes, it has been found that the proportion of connection in lowest shell presents obvious "V"-type fluctuations over time, which provides main basis of small fluctuation coefficient in the following chapter.Again, focusing on the hierarchy of mutation points, it has been found that difference of connection ratio turns to be cyclical, representing the "oscillation-smooth oscillation" law; mutation points are all in the transition of the fluctuations to the stable phase, the change of network occurs when small fluctuations accumulate to a certain extent. At that time, the structure of network adjusts, difference of inter-shell connection ratio narrows, and the network stabilizes then. In the process of self-growth, by metabolism and self-replication, network can maintain balance by resisting the changes of environment through macroscopic topology structure's mutation. But if the error of small fluctuation enlarges, the system will have a "giant fluctuation", which causes the mutation of characteristic quantities. After mutation, difference of connection ratio reduced, and the network stabilizes gradually.Next, concept of standard structure entropy which quantifies the stability of the network has been proposed. Entropy slowly increases over time, redundant connections in the network reduce, dependence weakens, and nodes tend to cross structural holes to reach out to non-redundant information source. Nodes with high degree cross more structural holes, more exposed to non-redundant information resources, occupying dominant position. Network transits to optimal and efficient state self-adaptively. Based on analysis of the five necessary conditions of dissipative structures, it can be concluded that the evolution of the Internet meets the formation of dissipative structures. Based on the conclusions of the first few chapters, the method of magnifying difference of connection ratio has been used to quantify the probability of mutation-small fluctuation. Using real topology data for experiments, the accuracy of the small fluctuation coefficient has been verified.Finally, advantages and disadvantages of the existing several classic models have been introduced and compared. Based on the PFP model, taking difference between PFP model and real network in the nuclear distribution into consideration, interactive growth of the new nodes and nonlinear preferential attachment modeling internal side has been carried on, connection and selection formula with growth and death node of the Internet practical Ark data has been fit, a new modeling method has been put forward. Simulation results show that a number of principle statistics in the model are similar to the actual Internet, not only maintaining power-law of the network, but also better reflecting core distribution and AS-level small fluctuation. The dynamic modeling algorithm can reconstruct the evolution of the Internet. Furthermore, it proved to be superior to the classic PFP model in links inside of shell.