Flex-grid technology increases network links capacity and optical power levels, creating power saturation problem in legacy amplifiers. We demonstrate that optimizing regeneration sites allows reducing the optical power of highly loaded links, avoiding amplifiers saturation over the existing fixedgrid networks.
IntroductionFlex-Grid is a promising technology that allows better spectral efficiency and spectrum usage in optical networks. It offers the possibility to create additional channels in the same available bandwidth (C band) and thus improving network capacity. However, this capacity increase may not be sustainable, because of the associated increase in optical amplification power 1 . In fact, the increasing number of optical channels gives rise to a power saturation problem in legacy amplifiers when migrating from fixed-grid to flex-grid networks 2,3 .We demonstrated in 2,3 that avoiding saturation problem is possible through adapting channels optical power. This adaptation is possible by converting the OSNR margins (OSN R margin ) into optical power attenuation over optical links 1,4 . However, this power saturation can always arise over highly loaded links as demonstrated in 5 , especially when there is no sufficient exploitable OSNR margins.In fact, the OSN R margin is an important parameter to optimize network performance and to increase its capacity as demonstrated in many works in the literature 6,7 . It gives the possibility to optimize different transmission parameters, such as modulation format, optical power, baud/bit rates and coding scheme. In this work, we consider only the channel power adaptation, in order to reduce the power level over highly loaded links and avoid power saturation. The higher is the OSN R margin , the higher is the power attenuation that can be applied to an optical channel 5 .In this paper, we will show how regeneration sites impact the OSNR margins of established channels. Then, we present through an example the execution result of the developed poweraware regeneration algorithm (PAR). Finally, the performance of the algorithm is evaluated.