Resource Optimization Technology In Hybrid Band Elastic Optical Network

With the rapid development of diversified bandwidth-intensive services,such as 5G,mobile internet and ultra-high-definition video,network traffic volume will continue to maintain a rapid growth.According to the white paper "Data Age 2025" released by International Data Corporation,the global data flow will rise to 163ZB by 2025.As an important part of the data traffic transmission and information infrastructure,optical fiber communication network will face huge pressure in bandwidth expansion.At present,even though the new elastic optical network can improve the efficiency of network resource utilization with the help of by fine-grained wavelength grid and flexible resource scheduling,the single-band optical network may be still with the limited spectrum range.It is difficult to provide more available bandwidth resources for service connection,and cannot meet the capacity requirements of future optical network development.Therefore,extending single-band to multi-band coexistence and constructing hybrid band elastic optical network will become the main way to improve the transmission capacity of optical fiber communication network in the future.However,hybrid band elastic optical networks introduce some new physical layer problems while improving network capacity.Specifically,with the expansion of the service band,the interval between optical channels is lengthened,resulting in nonlinear impairment with stimulated Raman scattering.Compared with conventional elastic optical networks,this impairment brings a new challenge to the signal-to-noise ratio(SNR)of optical signal transmission,and the constraints of network resource allocation will become more complex.Therefore,how to effectively reduce the deterioration of the optical transmission quality caused by impairment and improve the network resource utilization on the basis of constructing a new optical signal SNR evaluation model are the priority issues when designing resource allocation algorithms.Meanwhile,with the expansion of network service bands,the problem of inefficient allocation of conventional resources in elastic optical networks also urgently needs to be addressed to synchronously improve network capacity.With the dynamic arrival and departure of network service requests,the irregular occupation and release of network resources may change them into discrete fragments,reducing the availability of network resources.In hybrid band elastic optical networks,due to the expansion of service bands,the optimization of available resource allocation is more important.Therefore,how to develop a fragmentation avoidance scheme for dynamic services at the network layer and improve the availability of network resources are the key problems to be solved urgently to improve the network capacity.In addition,if we can effectively obtain the demand changes of dynamic services,and then formulate the corresponding resource allocation plan in advance,we can alleviate the problem of resource utilization inefficiency caused by uncertain traffic changes at the service layer.Therefore,how to build an accurate traffic prediction model based on the fuzzy traffic characteristics in the optical network at the service layer,and then guide the realization of efficient network resource allocation are also the key issues to optimize the utilization of network resources.In view of the above problems and challenges,this dissertation focuses on the research of hybrid band elastic optical network resource optimization technology,where the considered band is expanded from the traditional C scenario to the C+L hybrid scenario.Multiple resource allocation schemes are developed including multiple impairment constraints,spectrum fragment avoidance and fuzzy traffic prediction in the physical layer,network layer and service layer of the network,respectively.Finally,the performance verification is carried out on the elastic optical network simulation platform under the multi-band coexistence environment.The related research works have important reference significance for the wide application of hybrid band elastic optical network and efficient resource allocation.The main innovation achievements are as follows:(1)Aiming at the research of physical layer,focusing on the basic transmission quality assurance under multiple impairment constraints,the hybrid band elastic optical network system is designed.Considering the impact of optical signal degradation in the transmission process,the optical signal transmission impairment assessment model is bulit.Furthermore,given that the upgrade of the multiple equipment bands in the network needs to be carried out step by step,the transition stage of the priority upgrade of some equipment is the main focus in this part.In this case,the band attributes of the link on the path are analyzed,the band resource availability of the selected path is distinguished,and the impairment-reduced resource allocation algorithm with SNR reverification is proposed.According to the required bandwidth and duration of the service request,the algorithm uses the first-last-fit method to complete the time-frequency two-dimensional resource allocation,on the premise that the SNR threshold of the corresponding modulation method is not lower than the threshold.In addition,in order to ensure the SNR of the service optical path throughout its full life cycle,the SNR of the service optical paths whose links with new optical paths shall be rechecked one by one to ensure that the additional impairment impact will not cause the existing optical path to deviate from the basic transmission conditions after the establishment of the new optical path.The simulation results show that the proposed algorithm can reduce the network blocking probability on the premise of reaching the highest average SNR of the optical path.(2)Aiming at the research of physical layer,focusing on the resource allocation efficiency under the joint constraint of physical layer impairment and resource allocation,a link resource balancing scheme under multi-band coexistence environment is proposed.The scheme first gives the definition of resource availability in the network,and focuses on the analysis of the different resource consumption of different links on the path selected by the service request.It designs a link avaliable resource weight evaluation method and a hybrid band resource optimization algorithm based on link balance.The algorithm not only considers the impact of stimulated Raman scattering impairment in multi-band scenarios,but also introduces the resource cost distribution of the service request path.The existing links with higher load may be set with a higher weight to reduce the further pressure on the available resources of these links when establishing a new optical path.The performance simulation results show that the proposed scheme can effectively reduce network blocking probability and improve resource utilization.(3)Aiming at the research of network layer,focusing on the waste of bandwidth fragmentation that are free but unavailable resources,this dissertation innovatively proposes a service-driven fragmentation measurement method compared with the general network-level method.The purpose of this method is to highlight the resource consumption of the bandwidth required by the service on the currently selected path and the links adjacent to the path,establishing a fragmentation impact assessment caused by the service.Furthermore,based on this method,a service-driven fragmentation-aware resource allocation algorithm is proposed,which strives to avoid to bring spectrum fragments to the current path and adjacent links as much as possible when establishing a new lightpath,so as to achieve efficient utilization of available resources.The simulation results show that the proposed resource allocation algorithm can effectively reduce the fragmentation of the average path in networks,and further achieve the performance enhancement of the network blocking probability.(4)Aiming at the research of service layer,focusing on the inefficient problem of dynamic resource planning under fuzzy traffic,this dissertation proposes a spatial-temporal traffic prediction model,which is different from the traditional machine learning model that only relies on time series prediction.Starting from the spatial characteristics of traffic transmission between nodes,it concerns on the node relationships in the network topology with a graph structure,and prefers to establish a spatiotemporal two-dimensional traffic feature extraction model with the graph neural network and gated recurrent unit to realize high-precision traffic prediction.Based on this model,a node-oriented traffic prediction and scheduling scheme is proposed to improve the balanced utilization of network resources from the perspective of routing.Furthermore,a node-oriented slice reconfiguration scheme is proposed to achieve an on-demand resource allocation from the perspective of routing and resource allocation,improving the quality of network service.