Impact Of ROADM Intra-node Add/Drop Contention On Lightpath Blocking Performance

Reconfigurable optical add/drop multiplexer(ROADM) which can offer flexible reconfigurability have been a popular technique for building a large-scale all-optical switched network. Today, most ROADM can support the colorless and directionless features; however, the contentionless feature is still not popular because of high cost. In this thesis, we introduce the intro-node contention ROADM node architecture, in which the number of lightpaths that use the same wavelength and can be simultaneously added/dropped is limited by the add/drop contention factor C(which equals to the number of add/drop port banks). To estimate the lightpath blocking performance of this type of node architecture, we develop two analytical models to predict the lightpath blocking probability for a single node and for an entire network respectively. Under dynamic traffic load, for different node degrees and network topologies, we perform extensive evaluations and compare the results between the analytical models and simulations. The results show that the analytical models are effective in predicting lightpath blocking performance under different combinations of add/drop contention factor C, number of add/drop ports per bank T, traffic load distribution, and offered traffic load. We have also considered three different adaptive routing and wavelength assignment(RWA) strategies. Simulation studies show that the RWA processes aware of the availability of add/drop ports at the source and destination nodes can significantly improve lightpath blocking performance compared to the non-aware cases. Finally, we also observe two performance saturation trends:(1) When the number of add/drop ports per bank exceeds a certain threshold level, performance improvement will be marginal even if the number of add/drop ports per bank is further increased.(2) Increasing the add/drop contention factor C, the lightpath blocking performance can be greatly improved; however, the lightpath blocking performance will not be improved with further increase of add/drop contention factor C. Finally, we also find that when the number of add/drop ports per bank and add/drop contention factor C exceed a certain level, the performance of a contention ROADM is almost as good as that of a much more expensive contentionless ROADM.