Wavelength routing in All-Optical networks using full, limited and no wavelength conversion

Abstract

We study the effect of full, limited and no wavelength conversion on the blocking performance of the all-optical networks. We first formulate an exact model for linear tree networks for the case of full and no wavelength conversion. Using this model we show that the performance of full wavelength conversion, after certain a point, is worse than the performance of no wavelength conversion. This point, which we call crossover point decreases as the number of wavelengths is increased. Using this exact model, it is fairly difficult to analyze some larger networks. Therefore, to study the effect of fiiU and no wavelength conversion we use an approximate model and present the same results. This surprising result conveys that wavelength conversion can hurt and after the crossover point no wavelength conversion performs better than full wavelength conversion. The crossover can be avoided if optimal algorithms are used since we can always pretend that we don’t have the converters but the point is that many analyses of the effect of wavelength conversion assume certain suboptimal routing algorithms and take for granted that wavelength conversion can only help. The results obtained show that it can sometimes hurt under the same routing algorithms. We find that this crossover does not occur below 10% blocking level. However, most of the time we are only interested in having at most 2 or 3% blocking probability. At these low blocking levels full wavelength conversion outperforms the no wavelength conversion. Although full wavelength conversion is desirable because it decreases the blocking probability, it is difficult to implement in practice due to technological limitations. Moreover, all-optical wavelength converters demonstrated in the laboratory to date are, in general, only capable of converting to a limited range of output wavelengths for any given input wavelength. Therefore, an interesting design alternative is to have a limited-range wavelength conversion and quantify its advantages vis-a-vis no wavelength conversion and full wavelength conversion. For this, we develop an analytical model to study the effect of limited-range conversion for any network topology. Using this model we demonstrate that the performance improvement obtained by full wavelength conversion over no wavelength conversion can be achieved by using limited wavelength conversion with the degree of conversion being only 1 or 2. In two example networks we considered, for 2% blocking the carried traffic with d — 2 limited conversion was almost equal to the carried traffic for full wavelength conversion. Our new analytical model for limited wavelength conversion is much more accurate than the models developed hitherto for the case of no wavelength conversion. Comparison with simulations in two example networks show that the difference between the performance predicted by our approximate analytical model and simulations is in the range 0.001-0.003 for the blocking probability in the range 1-5%.