The Research Of Virtual Topology Embedding And Survivability Technology For Cloud Service

Cloud Service is a cost-effective delivery model that provides IT resources for tenant applications in cloud computing environments.There has been lots of challenge for a variety of applications of cloud tenant requests to provisioning physical resources in multi-tenant,reliability,resilience and isolation problem.Virtual Toplogy is the core problem to provisioning cloud service.Whether to satisfy the requirements of all tenants and how to assign physical resources for the virtual topology of cloud tenant requests is called virtual topology embedding problem.There are not only difficult to satisfy cloud service requirements but also lead to low resource utilization under the unreasonable resource assignments.Virtualization technology abstracts the physical resources into virtual logic resources,making the logical topology of resources no longer dependent on physical devices and their topology,and is a core technology of the virtual topology embedding for cloud services.In the cloud data center,the virtual topology is abstracted into the virtual resources topology by using virtualization technology.The virtual topology of multiple cloud tenants can be mapped into the same cloud data center physical network topology,and different types of virtual topologies are built into the physical network topology.However,how to embed virtual topology efficiently into the cloud data center infrastructure is still facing many challenges.It has been a preliminary exploration on the existing virtual topology embedding approach in the granularity of virtual resource allocation,the attribution of resource logic topology and survivability,which has obtained some achievements in the academics and industry.However,the virtual topology embedding approach oriented the features of cloud service still faces the problem of the reconfigurability of virtual topology,the survivability of virtual topology embedding and multi-objective optimization.There are still many restriction in the software defined virtual topology embedding based on heuristic approach,the survivable virtual topology embedding without backup resources,and the multi-objective optimization of virtual topology embedding for the resilient cloud tenant requests.According to the shortcomings of the existing methods,this paper has launched in-depth study for the cloud-oriented virtual topology embedding approach and the survivability,which mainly includes the software defined virtual topology embedding,survivable virtual topology embedding and multi-objective optimization of virtual topology embedding.The main works and contributions of this paper are as follows in details:(1)The research of virtual topology embedding model and optimization method for cloud serviceThere are various network topology of cloud tenant requests,and the virtual topology embedding method has some limitations on the virtual network topology and path diversity.Based on the virtual topology embedding model and the attribution of the network topology,a network-aware multipath virtual topology embedding method,called NMVTE,is proposed to improve the topology embedding efficiency and the revenue and cost ratio(R/C).This approach fully takes consideration of the degree and closeness of the network centrality as the virtual topology node embedding factor and the multipath link embedding method to improve the efficiency,benefits and R/C of virtual topology embedding,which is based on the virtual topology embedding model to analyze the topology properties.A large number of experiments have been used for evaluating the runtime,request acceptance rate and R/C of virtual topology embedding,which is better than other similar methods.(2)The research on survivable virtual topology embedding for critical cloud serviceFor a large of cloud tenant request,it is difficult to guarantee the reliability of each application service.Providing survivable features for mission-critical services is an effective measure to solve the usability of cloud service applications.A network-aware survivable virtual topology embedding method,called SVTE,is proposed,which uses the component risk assessment model to improve the survivability of the virtual topology by enhancing the virtual controller node embedding strategy,and optimize the nodes and links in the virtual topology location selection without using any backup resources.This method takes full account of the network delay and the non-disjoint path between the controller node and the forwarding device,using survivable factors to minimize the probability of loss of control paths.A large number of experiments have been used for evaluating the survivability and other metrics in the case of no additional backup resources,and the results show that this approach enhances the survivability of the virtual topology embedding almost without degrade other indexes.(3)Multi-objective optimization approach of virtual topology embedding for resilient cloud serviceThe existing method has been lack of the comprehensive consideration in complex factors for the resilience of cloud service.A novel virtual topology embedding method called CoVTE,is proposed,which considers and adjusts the proportion of embedding factors and obtain flexibly the various the virtual topology embedding.In order to achieve the flexible and reliable virtual network topology,the network centrality,network graph correlation attribution and resource fragment optimization strategy are adopted to optimize the multi-objectives of virtual topology embedding,including the logical topology of the virtual machine,the virtual topology of the virtual switch node,the virtual link,and the virtual controller node.This method has been tested by a large number of experiments,where network delay,embedding acceptance rate,the utilization of physical resources and system throughput are better than other similar methods.(4)The prototype implementation of cloud-oriented virtual network clusterThe advanced technology and method are adopted to build a cloud-oriented large-scale cloud computing environment under the physical cluster system of MilkyWay-2(TH-2).A cloud service delivery model,based on the Fat-Tree physical network topology and virtual topology embedding model,is effectively aim to the convenience for cloud tenants by using software defined resources.To solve the problem of software defined virtual topology,application software dependency and resilient service,cloud computing technology makes the traditional high performance computing center into cloud characteristics,and this paper,according the above research,has launched a preliminary exploration for the cloud system of HPC.Experiments show that the building technique,approximate to the performance of the physical nodes,enables the service model to be the on-demand characteristics of cloud and obtain high efficiency of virtual topology embedding.