Study On Optimizations For Networks Performance Based On The Self-similar Model

Model of network traffics is the fundament of network architecture design and performance analysis, the accuracy of measurement and the precision of description are very important for optimization and improvement of the network performance. A large number of measurements and statistical analyses show that the self-similar characteristic is ubiquitous in the networks regardless of the topological structures and types of service. The early analyses and evaluations for network performance based on the Poisson model were no longer applicable, and the self-similar model can accurately describe the characteristics of the network traffics. Firstly, the effect on the self-similar characteristics is analyzed by a leaky-bucket algorithm on basis of modeling and detecting to self-similar traffic, and a fuzzy control mechanism is introduced to improve the characteristics of the output traffic. The experimental results simulated by a computer show that the improved leaky-bucket algorithm has better performance to the self-similar property of output traffic flows and better self-adaptability to the buffer capacity and the variance fluctuation of input traffic flows in heavy load condition. Secondly, the QoS (quality of service) parameters are discussed with the FBM (Fractional Brownian Motion) as a self-similar traffic input model based on the analysis of the self-similar queuing performance, and the mechanism of bandwidth allocation and admission control is studied for quality of service under the self-similar network condition. The experimental results simulated by a computer show that the effective bandwidth is a better description of the characteristics of self-similar traffics, and the improved resource allocation and call admission control based on the effective bandwidth can reduce the network delay and delay jitter, and increase the network throughput.The investigation results in this paper show that it is necessary and feasible for the optimization of self-similar network performance. The work of this paper can prove groundwork for network architecture, protocol mechanism, algorithm design of the next generation networks (NGN) and has important theoretical and practical significance to design and application of network engineering.