Research On Deployment Scheme For Service Function Chains In Datacenter Network

With the rapid development of communication technology,the service requests from clients tend to be more diversified and show a trend of explosive growth,which put forward a higher requirement on the scale of network.The traditional Internet architecture lacks flexibility,which can't realize sufficient sharing of network resource when handling differentiated service requests.Therefore,it may need plenty of hardware resource to expand the network scale.Under the background,the technology of network function virtualization(NFV)comes into being.NFV can decouple hardware and software,which allows network function of hardware to be deployed on universal server in the form of virtual network function(VNF).Each service request is a service flow processed by multiple network functions in turn.NFV enables each service request to be seen as a chain set of ordered VNFs,i.e.,service function chain(SFC).To reduce operating cost,service providers tend to outsource the service requests of customers to datacenters in the form of SFCs.However,the existing related works are still difficult to maximize the resource utilization of infrastructure network.Therefore,we investigate the deployment scheme for SFCs in datacenter network to optimize network resource utilization.For the problem of online SFCs deployment,we design a resource-efficient deployment method.Firstly,a cluster model of datacenter network is established to directly reflect the length of each virtual link's optimal mapping path.Based on the model,we expect to achieve a tradeoff between computing resource usage and bandwidth resource usage with the formulation aim of minimizing the normalized product of these two kinds of network resource usage.Afterwards,we design a heuristic algorithm,called as CBRT,to complete the SFCs deployment respectively in the scenario with sufficient bandwidth resource,the scenario with a serious lack of bandwidth resource and the scenario where bandwidth resource is relatively moderate to computing resource.The simulation result shows that our method of online SFCs deployment can adaptively achieve the full sharing of instance resource and obtain better network resource utilization in various scenarios of infrastructure network resource.For the problem of offline SFCs deployment,a method is designed to consider resource reservation when deploying SFCs.We expect to activate fewer servers when completing the offline deployment for a series of known SFCs.Meanwhile,we consider reserving a certain amount of resource on the activated servers so that subsequent SFCs deployment can share the existing instantiation resource.In this thesis,we formulate the deployment target as a mathematical model,which is called DMCRR.Afterwards,a corresponding heuristic algorithm is designed to solve the model,which consists of two stages.One is initial deployment,which requires the minimum computing resource consumption.The other is VNFs adjustment,aiming to reduce the load level of partial physical links.The simulation result shows that our method can not only efficiently complete the offline deployment of known SFCs,but also obtain better subsequent deployment capacity.In summary,the method can optimize overall network resource utilization to a greater extent.