Network Traffic Model Based On Fractional Differencing And Fractal Filter

Along with the rapid development of the computer network, the current network on an epic scale and complex, based on network applications increase sharply, network interconnection environment is more complex, means that the more problematic Internet service, network performance is more susceptible.To give the user to provide high quality service, network maintenance and management is particularly important.Network traffic is accurate prediction in computer network design and management, conflict control and dynamic bandwidth allocation plays a very important role.However, successful flow prediction without precise flow model of support, high quality flow model for designs high-performance networking protocols and effective network topology structure, To design a cost-effective network equipment and the server.For accurate network performance analysis and prediction: to congestion management and flow equilibrium improve service quality has very important significance, accurate network performance analysis and prediction without accurately network flow model.The essay analyses the current network flow generally exist at long related and short relevant features, discovered that these two features of network flow trend and other aspects has a different effect.In a small granularities representation of flow sequence, the shorter related process of network flow influenced, as time increases granularity of long related and short related will also effect on the performance of the network, until the scale is big enough, flow sequence of short related process almost no longer effect, a leading long correlation.Using this variable-dimension nature difference characteristics in different granularities of network flow of different processing, will be targeted raising network monitoring performance and effect.Based on the current various network flow model of research and analysis, find can also describe network traffic short and long related process related FARIMA model, this model the process of decomposition, obtained the related process for long FARIMA (0,0), d, namely score difference process.To score difference process carried out a series of analysis and simulation experiment after detection, scores difference can effectively weaken network flow sequence of long related structure, thus corresponding get short related characteristics mainly smooth sequence.Using this idea, analyzes the inverse process score difference fractal filtering, it is found that the fractal filter, can strengthen the time series of long related structures.This structure network traffic variable-dimension nature difference characteristics, was put forward based on the score difference and fractal filtering network flow model. Application of large time of particle size of network traffic statistical records, to obtain the time sequence of fractal filtering operation, enhance the sequence of long related structure, application suitable traditional long related model, thereby obtain larger modeling predict future time span flow trend; In network anomalies frequent peak stage, the application as little time of particle size of network traffic statistical records, to obtain the time series score difference operation, weaken the sequence of long related structures, difference sequence application suitable traditional short related model to predict the network modeling forecast, abnormal happen.Through the simulation test, this method is compared with the traditional flow model to exist at the long related and short related network traffic has better prediction effect.