A contribution to optical packet switching: Switching architectures, performance evaluation and comparative analysis

Application of Optical Packet Switching (OPS) to the transport WDM network is observed by network operators as a promising alternative, in particular when photonic devices technology achieves the required maturity level. This work focuses on the performance evaluation of OPS switching fabric architectures based on fiber delay lines, which is the critical component in network cost. Architectures are classified attending to the network operational mode: SHWP (Shared Wavelength Path) and SCWP (Scattered Wavelength Path), which are the competing candidates for the control of OPS networks.; The work is structured in three sections: (1) Output Buffering (OB) OPS switch fabrics. A set of architectures were identified. An optimum scheduling algorithm was proposed for SCWP operational mode, which cleanly outperforms the SHWP version, also maintaining packet sequence. Queue analysis were proposed which provide the exact packet loss probability, exact packet delay distribution, and exact busy period distribution. This research has allowed the dimensioning process, and a cost comparison among different OPS OB architectures. (2 )IB-WR (Input-Buffered Wavelength-Routed ) switch. The interest in this architecture stays on its minor cost and better scalability. First contribution is the demonstration that the per-time-slot scheduling problem can be expressed as a matching problem in a bipartite graph. This novel formalization builds a bridge between this research, and the research in electronic VOQ (Virtual Output Queueing) switches. We have studied the similarities and differences between VOQ/IB-WR optimization problems, to take benefit of the existing research in fast VOQ scheduling algorithms. The result has been the proposal of the PDBM (Parallel Desynchronized Block Matching) IB-WR scheduling algorithm for SCWP switches. This is the first algorithm of this type in the literature. The performances of this algorithm has shown to be extremely close to the optimum achievable, given by OB architectures. (3) OPS large-scale switching fabrics. The previous proposals have been reviewed and adapted to SHWP/SCWP operational modes. Also, a novel 3-stage OPS architecture, based on the knock-out effect, has been presented. The mathematical analysis for knock-out packet losses evaluation has been provided. Finally, a full cost and performance comparison of large-scale proposals have been evaluated for both operational modes.