We study the configuration issue of three-stage multi-granularity optical cross-connects (MG-OXC) for the dynamic traffic model in all-optical networks. From the single node point of view, we propose a configuration algorithm to configure different granularity cross-connects for arrival sub-requests with different traffic types and bandwidths. The performance of the configuration algorithm is evaluated by simulation and, furthermore, is validated by experiment based on our flexible Multi-functional Optical Switching Testbed (MOST).

We develop a new optical switching fabric with multi-granularity grooming based on our lambda-group model, as well as algorithms that can handle dynamic environments. The proposed fabric based on a new multi-granular grooming scheme presents the distinctive approach of assigning different contiguous groups of granularities to different paths for effective treatment. Results and figures from experiments show that the particular partitioning approach not only is helpful to port reduction significantly, but also improves the SNR of signal and blocking performance for dynamic connection requests.