Research On Resource Scheduling Algorithm And Experimental Platform For Cloud-network Integration

The future of networking needs to meet the needs of emerging business services such as industrial internet and internet of vehicles,and needs to overcome the challenges of flexibility,security,and other issues.Therefore,the future network technology is booming.At present,software-defined networking(SDN),network function virtualization(NFV),and content centric networking(CCN)have become the research hotspots in the future network.The software-defined networking realizes the separation of forwarding and control,the network function virtualization realizes the software and hardware decoupling of the network element function,and promotes the realization of the software and virtualization of the network.The openness and programmability of the network can better promote the innovation of network business,and also enable the simple and efficient operation of the infrastructure.Cloud network integration is an important development trend of the future network.It integrates the use of cloud computing,software-defined networking,and network function virtualization technologies to achieve effective integration of cloud network resources.The intelligent,converged information and communication infrastructure centered on the data center has become the mainstream in the industry.Cloud-based architecture enables flexible allocation of computing,storage,and network resources,while enabling flexible layout management based on the diverse needs of the business.The use efficiency of the resource can be effectively improved through the cooperative management of cloud network.This dissertation focuses on the challenges of intelligent,co-scheduled scheduling of resources such as computing,storage,and networking in the future network.It focuses on solving technical difficulties such as virtual machine migration and integration scheduling,dynamic mapping of service chains,content cache placement,and experimental platform design and development.The specific research content and main innovative work include the following:(1)Proposing a virtual machine integration algorithm based on workload prediction.Under the background of cloud network integration,virtual machines are migrating more and more frequently due to load balancing,automatic scaling,green energy saving and disaster recovery.This dissertation focuses on the implementation of virtual machine integration scheduling based on workload prediction.A new virtual machine hybrid control system based on active control technology based on workload prediction and passive control technology based on actual system state information is designed.By using an exponential smoothing prediction model to predict the workload of the virtual machine in the future,a virtual machine integration algorithm is proposed to compare the maximum future workload priority in the virtual machine selection phase and the resource demand queue in the virtual machine placement phase.Simulations show that the algorithm reduces resource usage,virtual machine migration times,and service level protocol violations using a prediction-based resource consolidation approach.(2)Proposing a service chain mapping algorithm based on reinforcement learning.The mapping of service chain is an important technical means to instantiate the network business in physical infrastructure,and it mainly realizes the dynamic provision of network service resources.This dissertation designs a service chain resource scheduling architecture based on multi-agent reinforcement learning technology,and builds a multi-agent intelligent collaborative management system by adding new intelligent service planes to manage the resources of multiple network domains.A service chain mapping algorithm based on Q-Learning reinforcement learning is proposed.The Q-Learning mechanism is used to calculate the deployment location of each virtual network element in the service chain.The algorithm effectively reduces the average transmission delay of the service and improves the load balance of the system.(3)Proposing a content pre-caching mechanism based on bipartite graph.The integrated storage function in the network will be an important development trend of the future network.It is necessary to re-consider the mechanisms of user request scheduling,network resource allocation,and cache update.This dissertation designs the integration of storage capabilities in network device nodes such as mobile network base stations to improve the efficiency of content distribution.A bipartite graph-based mobile network content pre-caching strategy is proposed.The cache node is selected by considering the popularity of each base station node and the network link status between nodes.This strategy greatly reduces the delay of content transmission,improves the hit rate of cache nodes in the-network,and is more conducive to the efficient distribution of content.(4)Designing a future network experiment platform based on SDN/NFV.Domestic and foreign large-scale experimental beds have been built and related experimental management systems have been built.This-dissertation gives full play to the technical advantages of SDN/NFV,uses SDN’s centralized control delivery flow table to implement cross-domain Layer 2 and Layer 3 communication and link service quality assurance,The experimental platform implements standard definition and life cycle management of virtual network elements based on NFV,using flexible scheduling service chain to realize the network business.The platform realizes the on-demand provision of the underlying physical infrastructure resources through unified abstract scheduling.