Research On Resource Scheduling Algorithms Based On Node Architecture On Demand In Space Division Multiplexing Elastic Optical Network

Compared with wavelength division multiplexing networks,elastic optical networks(EONs)can flexibly allocate spectrum resources according to the bandwidth of the service and thus can reduce the waste of spectrum resources and improves resource utilization.However,with the increasing growth of network traffic and the continuous expansion of network scale,fibers with single-core and single-mode deployed in EONs at the present stage will not meet the capacity requirements of future networks due to Shannon's limit and will seriously affect the development of future networks.Multi-core fiber is an effective technology to overcome the capacity bottleneck of current single-core and single-mode fiber.Space division multiplexing elastic optical network(SDM-EON)enabled by multi-core fiber has been regarded as a potential next-generation underlying transmission network.Although SDM-EON has many advantages,its development faces new challenges.First,in a switching fabric,the overlap of spectrum requested by different requests will cause contention,resulting in waste of node device resources and spectrum resources.Second,core-switching makes the resource allocation problem more complicated.In addition,inter-core crosstalk(XT)between adjacent cores in multi-core fiber affects the availability of spectrum resources and the transmission quality of the service,resulting in an increase in the network blocking probability and restricting the development of the network.Based on the analysis above,node scheduling algorithm and routing,spectrum and core allocation(RSCA)considering XT are taken as two research aspects of the thesis.The main work and innovations of the thesis are listed as follows:Firstly,in order to solve intra-node contentions in SDM-EON,a switching fabric configured with spectrum converters which is based on architecture on demand(AoD)is designed,and the space-frequency joint scheduling algorithm(SFJSA)based on weighted maximal clique is proposed.If conflicts exist among requests,an auxiliary graph is created,the requests and the conflict relationship between requests are abstracted as the vertices and edges of the auxiliary graph,respectively.Then space switching is performed to solve the conflicts.If contentions still exit after space switching,spectrum conversion will be performed.In the stage of space switching,vertex weight formulas based on the core loads and the number of used spectrum selection switches(SSS)are designed.In the stage of spectrum conversion,vertex weight formulas are designed based on thebandwidth of the request and the spectrum compactness after conversion,respectively.The maximal clique with the maximum weight is the scheduling scheme with the maximum output bandwidth and the minimum space switching/spectrum conversion costs.The simulation results show that the proposed SFJSA can reduce the number of required spectrum converters while improving the node bandwidth blocking probability(BBP)performance,achieve better performance in a more cost-effective manner.Secondly,in order to reduce inter-core crosstalk in multi-core fiber based SDM-EON,a RSCA algorithm considering node spectrum conversion capability and XT(NSCC-XT-RSCA)is proposed.Spectrum converters are adopted to optimize the crosstalk in the proposed algorithm.Considering the high costs brought by full distribution of the node capable of spectrum converting,the sparse distribution based on the betweenness centrality of nodes is adopt,the nodes with high value of betweenness centrality are chosen to configure spectrum converters.In the stage of routing,a path weight formula based on the path loads and node spectrum conversion capability is designed to find paths with lights loads and high spectrum conversion capability.In the stage of core and spectrum allocation,the cores are grouped based on vertex coloring and the spectrum is partitioned according to the number of core groups.Based on the result of core groups and spectrum partition,a formula is designed to measure the costs of core and spectrum allocation,the allocation method with lowest costs is selected.In the stage of optimizing the crosstalk,spectrum conversion will be performed if the request does not satisfy the crosstalk threshold.Considering the case of multiple nodes on the path can perform spectrum conversion,a node ranking formula based on the number of remaining spectrum converters and the crosstalk improvement rate is designed to select the node.The simulation results show that NSCC-XT-RSCA can reduce the blocking probability caused by XT,reduce BBP and improve the resource utilization of networks.