Research On Optical Network Optimization Design For Asymmetric Traffic

With the popularity of new technologies and applications such as data centers and Internet of things,the Internet traffic has been in an explosive growth and network traffic asymmetry between upstream and downstream has become increasingly prominent.However,almost all of the current network designs are based on the assumption that bidirectional traffic demand is symmetric.Using bidirectional symmetric design to deal with highly asymmetric traffic will inevitably lead to large wastage of network equipment and spectrum resources.Thus,it is important and meaningful to develop a new network design approach that can handle the traffic asymmetry.In this thesis,we propose a new network design approach aiming at the asymmetry between upstream and downstream traffic for an Elastic Optical Network(EON).First,we propose a unidirectional network design approach for traffic asymmetry.Compared with the traditional bidirectional design,the unidirectional design can allocate network resources efficiently according to the actual traffic flow,thereby significantly reducing network hardware cost and spectrum usage.We further consider the strategy of separating the bidirectional transponder,which decouples the bidirectional transponder into a pair of independent unidirectional transmitter and receiver.With the consideration of different types of transmitters,we propose two optimization models and corresponding heuristic algorithms.The results show that the separating transponder strategy can effectively reduce the number of network devices without sacrificing the spectrum efficiency.At the same time,the heuristic algorithm is very efficient,and the results of algorithms are very close to that of the optimization models.This thesis further evaluates the influence of the spatial traffic asymmetry on network performance.In addition,we also evaluate the proposed unidirectional network design scheme in the context of multi-period network planning.The results show that,with an increasing number of data centers in a network,which corresponds to a decreasing spatial traffic asymmetry,the network cost decreases.For the multi-period network planning,the strategy of decoupling a transponder can efficiently reduce the network hardware cost.