Research On Cross-layer Resource Flexible Scheduling For IP+optical Cooperative Network

With the comprehensive cloudification of network data storage,the rapid development of Internet services,and the continuous increase of rich new services,the trend of IP-based network services is more and more obvious.With this trend comes the ever-expanding network scale and the growth of various heterogeneous dynamic applications.In this situation,IP packet service often presents the characteristics of high dynamics,high burstiness and high peak-to-average ratio,but at present,the backbone network adopts IP+optical cascade architecture,under this architecture,IP layer and optical layer are under hierarchical control and independent operation and maintenance.There are some problems,such as insufficient flexibility,low resource utilization,insufficient reliability,difficult control and maintenance,high investment and operation and maintenance cost.How to effectively cooperate the IP+optical network to flexibly schedule network resources carrying services across layers,reduce operation and maintenance costs,improve network resource utilization and network service quality,and reduce energy consumption and cost as much as possible is worthy of in-depth study.Therefore,in view of the above problems,this paper mainly studies the control architecture,resource optimization allocation mechanism and delay optimization control mechanism.1?Based on the core idea of SDN,the collaborative control architecture of IP+optical cross-layer network based on resource pooling is designed.In order to solve the coordination problems caused by hierarchical control of IP and optical network,as well as the low efficiency of resource utilization,lack of flexibility,complex maintenance and control,cumbersome collaborative scheduling process and long cycle caused by different control interfaces and private protocols of equipment from different manufacturers.This paper designs a four-layers hierarchical management and control architecture based on SDN to monitor the IP router port resources and optical link resources globally,and flexibly manage and configure the devices through the controller,so as to form a network resource pool,realize flexible resource scheduling,and then realize the network following the cloud and efficient resource utilization.Finally,the feasibility of the scheme is verified by building a hardware experimental platform.2.A cross-layer network flexible adjustment and resource allocation mechanism based on resource pooling was proposed.In view of the low utilization rate of network bandwidth resources and high energy consumption caused by the high dynamics and peak-to-average ratio of emerging services,this article deeply analyzes the relationship between services traffic and network resource utilization,network energy consumption under the IP+optical network model,when the network bandwidth is allocated by different resource allocation methods.Established a staged optimal resource allocation model for energy consumption based on a schedule in a multi-business scenario.At the same time,under the premise of dynamic adjustment of the cross-layer network,a real-time multi-service cross-layer resource allocation algorithm(MS-SNA)based on genetic algorithm is proposed.According to the input of the multi-service traffic matrix and the maximum tolerable delay,the algorithm plans the lowest energy consumption network resource allocation plan according to the timetable.And carry out the configuration of IP network and optical network to adjust network resources in real time,achieve the network resources efficiently with the lowest energy consumption,and ensure the quality of services.A large number of experimental simulations verify that the proposed algorithm can effectively use network resources and reduce the services congestion rate and energy consumption.3.A multi-period network adaptive reconfiguration mechanism based on optical path pre-configuration with machine learning and resource delayed release was proposed.Aiming at the problems of the fragmentation and low utilization of optical spectrum caused by the dynamic and high burstiness of network services,and the long service delay caused by the demolition and construction of optical paths for the effective use of optical resources,this article analyzes the relationship between the establishment of optical path and the services provide delay in depth.Established periodic optical path adaptive reconfiguration and spectrum sorting models under different service traffic models,and proposed multi-period network adaptive reconfiguration based on optical path pre-configuration with machine learning and delayed release of resources.The algorithm uses the long and short-term memory model of deep learning to train historical spectrum allocation data,and predicts the spectrum allocation plan of each link to perform optical path pre-configuration to reduce service delays and improve spectrum utilization.At the same time,the strategy of delayed release of hotspot optical path resources is combined to cope with the delay introduced by the newly-built optical path under abnormal burst traffic so as to optimize the control of the delay.Simulation experiments verify that this algorithm can effectively improve the spectrum utilization and reduce the average services provide delay of the services under different traffic loads and network models.