mited range wavelength conversion. Eased on the results obtained, we conclude that the proposed analytical modells simple andyet caneffectively analyze the impact of wavelength conversion ranges and number of fibers onnetwork perfomlance. Also a new hmtristic approach for placement of wavelength converters to reduce blocking probabilities is explored Finally, we analyze network peffomlance with the proposed scheme. It can be observed from numerical simulations that limited range converters placed at a few nodes can provide ahilost the same blocking probability as full range wavelength converters placed at all the nodes. We also show that being equipped with amulti-fiberper-link has the same effect as being equipped with the capability of limited range wavelength conversion. So a multi-fiber per-link network using limited range wavelength conversionhas similar blocking peffomlance as a full wavelength convertible network Since amulti-fiber network using limited range wavelength conversion couldusefewer convertersthan a single-fiber network using limited range wavelength conversion and because wavelength converters are today more expensive than fiber equipment, a multi-fiber network in condition with limited range wavelength conversion is less cosfly than a single fiber network using only hiltited range wavelength conversion. Thus, multi-fiber per-link network using limited range wavelength conversion is currently a more practical method for all optical WDM networks. Simulation studies carried out on a14-nodeNSFNET, a10-node CERNET (China Education and Research Network)，and a 9-node regular mesh network validate the analysis.