Most, if not all, theoretical work on the capacity design of mesh-restorable networks has been done in an integer, but non-modular, fashion. Although OC-n modules can only approximate the exact design capacities, this has been acceptable in practice where networks are rapidly growing and demand forecasts have uncertainty in any case, or in research studies where modularization only confuses the comparison of underlying theoretical effects. However, the modularity and economy-of-scale effects in near-term dense wave division multiplexing (DWDM) transmission systems may be so large (consider 40 /spl lambda/ at OC-192 on a single fiber) that we need to study and exploit these effects in the optimal capacity design problem. To do so, we introduce and compare results with two approaches for modular optimal capacity design to non-modular reference designs which are subject to conventional post-modularization. Significant shifts in the network structure, and total costs savings up to 20% are seen in the test results.
展开▼