| Optical packet-switched networks, the most promising infrastructures to support data traffic streams such as IP, are widely investigated, because they are scalable, efficient in bandwidth utilization, transparent to data rate and data format, and daptable for various traffic patterns. As the interfaces between standard electronics networks and optical core networks, edge nodes implement optical packet assembly, optical packet header insertion, traffic shaping and other functions. It has a significant impact on the whole network performance. As far as significant characteristics of data traffic are considered, this dissertation makes a thorough research on traffic source modeling, network traffic characteristics analysis, optical packet assembly mechanism and queue strategies in edge nodes by computer simulation, numerical calculation and theoretical analysis.After a concise survey of the progress and the key technologies of optical packet switching is presented, two self-similar traffic source models, superposition of fractal renew process (SupFRP) and fractal-binomial-noise-driven Poisson process (FBNDP), are built to approximately generate actual traffic data in Chapter 2. Weighted average autocorrelation function (WAACF) and unbiased wavelet estimator are adopted to quantitatively investigate short-range dependence (SRD) and long-rang dependence (LRD) of self-similar traffic data generated by these two models, respectively. Conclusions are drawn as follows: the deviations in high and low Hurst value are caused by both the strong effect of SRD and the limitation of models, respectively; the optimal Hurst value exists in the range of 0.7 to 0.85; the peak rate R should be given a medium value for FBNDP.In chapter 3, a novel method to estimate the degree of self-similarity based on the index of dispersion for counts, IDC estimator, is proposed. After making a performance comparison among R/S estimator, wavelet estimator and IDC estimator, we obtain that...
|
PDF Full Text Request