Research On High-availability Network Function Virtualization Resource Allocation Mechanism

With the development of the Internet,the network carries more and more applications and services as the communication base,which brings new re-quirements and challenges to the traditional network.Therefore,a variety of network middlewares are added to the network,which makes the traditional network complicated and rigid.Network Function Virtualization(NFV),as a promising technology,deploys network functions as software in the underlying network to replace network middlewares.NFV decouples network functions from hardware devices,bringing greater flexibility to the network.In NFV,VNFs(Virtual Network Functions)provide services in the form of SFC(Ser-vice Function Chain).Under various demands,how to place VNFs and SFCs on the underlying network(i.e.,resource allocation problem),has received in-creasing attention from both academia and industry.Besides,as an important part of QoS(Quality of Service),availability is also widely concerned in the NFV resource allocation problem.Compared with traditional IT applications that only require availability between two 9s to three 9s(i.e.,99%and 99.9%),telecom network services usually require five or even six 9s.Therefore,how to effectively improve service availability is an important and urgent problem to be solved.In recent years,the NFV resource allocation problem has received exten-sive attention and many outstanding results have been achieved.However,there are still some problems remaining to be solved,such as availability guarantee in data center networks(including VNF availability guarantee and parallelized SFC availability guarantee),VNF availability guarantee and SFC placement in mobile edge computing,resource modeling and resource allocation problems in network slicing,etc.Therefore,this paper studies the high-availability NFV resource allocation mechanism.The contributions are as follows:·This paper proposes an VNF placement method that combines availabil-ity guarantee and resource optimization in the data center network.This method defines an availability model that considers both physical device failures and VNF failures;proposes a novel joint backup model,which combines two existing backup solutions;designs an affinity-based algo-rithm to place VNFs.The proposed solution can optimize physical link and physical node resource consumption in the underlying network while guaranteeing availability.·This paper proposes a parallelized SFC placement method that guarantees availability in a data center network.This method proposes a multi-flow backup model to guarantee the availability of parallelized SFC;proposes three placement strategies and a hybrid placement algorithm based on two existing NFV execution models to optimize physical link consumption.The proposed solution can effectively reduce transmission delay,guarantee availability,and optimize physical link resource consumption.·This paper proposes an efficient SFC placement method that guarantees availability in mobile edge computing.This method formulates the SFC placement problem in mobile edge computing as a weighted graph match-ing problem,including two sub-problems(i.e.,graph matching and SFC mapping);proposes a graph matching algorithm based on linear program-ming and a placement algorithm based on Hungarian to solve the above two sub-problems.Besides,an availability matrix is used in the placement algorithm to guarantee availability.The proposed solution can efficiently execute in the mobile edge computing environment to optimize resource utilization of physical links and physical nodes.·This paper proposes an availability-aware cloud-edge collaborative net-work slicing resource modeling and allocation method.Cloud edge col-laborative network slicing defines a series of descriptors to describe the resource constraints and performance requirements of network slicing;propose a placement algorithm based on particle swarm optimization to solve the cross-domain(cloud and edge)resource allocation problem in cloud-edge collaborative network slicing.The proposed solution uses de-scriptors to abstract and model resources to simplify service descriptions(i.e.,placement constraints and availability requirements).The placement algorithm in resource allocation can parse descriptors and reduce resource consumption.