Research On Traffic Grooming Algorithms For Multi-Granularity Optical Switching

With the rapid development of Internet technology and the rapid growth of datatraffic, the study of WDM optical networks is greatly promoted. As the WDMTechnology becomes more matured, the obstacles for further development of the opticalnetwork transmission capacity are no longer the fiber bandwidth, but the bottleneckexerted by the equipment switching. However, the proposal on the traffic groomingtechnology and the technology of Multi-Granularity Optical Switching can easily getthis issue solved. Therefore, it is significant to study the traffic grooming algorithm inthe Multi-Granularity Optical Switching.This paper focuses on the research on the traffic grooming in the Multi-GranularityOptical Switching, as well as the analysis on the dynamic traffic grooming algorithm.This paper proposes different traffic grooming strategies for different traffic models byusing auxiliary graph models.The contents of this dissertation are listed as followed:(1) In the dynamic traffic model, this paper proposes a simplified KSP dynamictraffic grooming algorithm which removes overlapping paths. It can efficiently takeadvantage of the network resources and further reduce the blocking rate of traffic byusing a novel auxiliary graph model.(2) In the static traffic model, typical series of waveband sizes and differentperformances of waveband sizes affected by the K-means clustering algorithm areanalized. The relationship among the traffics in the Multi-Granularity Optical Switchingis studied and the appropriate waveband sizes are finalized.(3) With the novel auxiliary graph model based on waveband swithing, theproblems on waveband switching in the Multi-Granularity Optical Swithing areanalized, and different strategies on waveband traffic grooming in the static trafficmodel are proposed. The Analysis suggests that the performance of the traffic groomingstrategy by using the K-means clustering algorithm could be improved to some extent.This traffic grooming strategy not only makes full use of wavelength resources, but alsogreatly reduces the costs of network.