Research On Delay Sensitive Resource Allocation In Network Function Virtualization Environment

In recent years,with the rapid development and rapid update of network technology,the connection between people's daily life and the network has become closer.The Internet dramatically facilitates people's daily activities such as consumption,learning,and entertainment.Various emerging network services,such as e-commerce,live entertainment,short video,smart home,VR/AR,and smart home,have also brought new challenges to current network resources,bandwidth capacity,and latency.In wireless access networks,5G technology is developing rapidly and is gradually marketized;in the field of core networks,cloudification of network services is a general trend,so network technologies such as cloud computing and edge computing are showing their strengths in all aspects.The future network needs to improve the convenience and flexibility of management,reduce the cost of network construction,and improve the monitoring and scheduling capabilities of resources,such as allocating network resources on demand.Therefore,network function virtualization technology is one of the critical technologies for building next-generation networks.The emergence of network function virtualization(NFV)technology enables a large number of network functions originally based on dedicated physical facilities to be implemented on standardized general-purpose equipment in the form of virtual network functions(VNF)in software.These VNFs allow network operators to flexibly deploy at various locations in the network and migrate between different network nodes.Therefore,NFV technology can help network operators significantly reduce network construction and operation and maintenance costs.The unique software properties of the VNF itself,such as programmability and portability,also improve network resource management efficiency.In a cloud network that supports NFV,network service requests from users are executed by a Service Function Chain(SFC),and each SFC is composed of a series of ordered VNFs.Correspondingly,a user's request to use an SFC is called a service function chain request(SFCR),and each element in the SFCR is called a virtual network function request(VNFR).When the service provider receives a batch of users' SFCR,it needs to place and instantiate the corresponding VNF to complete the service request.Considering the unique software nature of the VNF,this allows service providers to choose the location and allocate resources for it flexibly.Simultaneously,to minimize the additional bandwidth overhead when mapping SFCR,it is also necessary to consider the order constraints between the VNFs involved.Therefore,in mapping SFCR and placing VNF,it is essential to formulate an appropriate plan to optimize network resource overhead and improve resource utilization.This type of problem is usually named after the VNF placement problem.The VNF placement problem,as a significant branch of the virtualization-based resource scheduling problem,has received extensive attention and research in the industry and academia.Reducing resource expenditure and network resource utilization in the cloud network while guaranteeing user service requirements,thereby reducing network operators' operation and maintenance costs,is a practical and challenging problem.This paper will focus on the QoS-aware VNF placement problem in a virtualized environment.Based on different optimization goals and scenarios,the research of this article mainly focuses on the following three aspects:1.We formulate the resource-efficient and traffic-aware VNF placement problem as a jointly constrained optimization problem,considering the time-varying arrival rates of users' SFCRs.Then we introduce J-ORTC approach which aims to minimize the total incurred resource and network traffic consumption in the system.To derive the solution for J-ORTC problem,we propose an algorithm which is based on biogeography-based optimization(BBO)algorithm.Since the original BBO is expensive to directly solve JORTC,we carefully seek a penalty function that reduce the original BBO" s state space of feasible solutions.The combination of the penalty function and BBO is EBBO algorithm,which efficiently reduces the convergence time.We then compare J-ORTC approach with other approaches in several case studies.Our simulation results demonstrate that J-ORTC approach provides near-optimal solutions and effectively reduces the total incurred consumption in the system.Furthermore,J-ORTC achieves lower total consumption than the benchmarks.2.We take FROs and the sharability of VNFs into account and reveal the conflict among node resource consumptions,delay constraints and band-width consumptions.And we leverage the feature that VNFRs on the shareable VNF instances can share FROs to save more instances,thus achieving a balance among these optimization objectives.We involve these two factors into our problem and formulate it as an ILP model.Then we propose the novel SFCR mapping algorithm(SMA)to map SFCRs to the nodes while guaranteeing their delay requirements.Moreover,we also devise the VNFR adjustment algorithm(VAA)to save more node resources.The evaluation results show that our approach can obtain a near-optimal resource consumption when the number of SFCRs is small.Besides,with a large number of SFCRs,our approach outperforms the benchmarks in terms of resource consumption,average VNF utilization,and the number of activated nodes.3.We consider fundamental resource overheads(FROs)and the sharing mechanism of VNF instances,which are two factors neglected in most related works.Utilizing the feature that VNFRs on shareable VNF instances can partake FROs,we can consume less server resource while achieving a balance between server resource and bandwidth consumption.Then we mathematically formulate the resource optimization and delay guarantee VNF placement problem.To solve the problem,we devise a polynomial-time approach called TPOS.In TPOS,we consider releasing redundant VNF instances based on the time-varying network load,which can save more server resources and improve resource utilization.We conduct a detailed analysis and evaluation of TPOS.The evaluation results show that TPOS can obtain a near-optimal server resource consumption when the problem scale is small.For large problem scale,compared with contrastive approaches,TPOS can effectively guarantee the delay requirements of SFCRs and outperform benchmarks in terms of server resource consumption,activated servers,and serve resource utilization.