A Research On Orchestrating Virtual Network Function Deployment Under Resource Constraints

The service provided by the traditional mobile communication network is based on the deployment of dedicated physical devices by network operators for each specific network function.Cellular core as a critical piece of the mobile communication network infrastructure,its centralized architecture leads to high coupling between physical com-ponents.Meanwhile,various network functions rely heavily on dedicated hardware,which leads to poor scalability and high management complexity.Furthermore,with more and more mobile connected terminals,increased deployment of physical devices always incurs high operating costs.The emerging paradigm of Network Function Vir-tualization(NFV)and Software Defined Network(SDN)bring new opportunities for communication operators.Decoupling software from dedicated hardware and function ion makes it possible to flexibly deploy service chains on commodity servers and fine-grained control the routing policies in a centralized way.Flexible and scal-able network architecture is conducive to the rapid iteration of new business and the reduction of operating costs.Nowadays however,there still exists many challenges on the deployment of Vir-tual Network Functions(VNFs)in the network.In view of the deficiencies in the current related research work,this paper proposes a research on orchestrating virtual network function deployment under resource constraints.Our goal is to minimize VNF deploy-ment cost.The main contents of this paper are divided into the following two parts:1.In the first part,we present a two-stage optimization framework of VNF deployment under resource constraints called Plutus.The first stage is network-level optimiza-tion,it aims to minimize the VNF service chain deployment cost.While server-level optimization is the second stage,it requires to determine which VNF should be deployed onto which CPU core to balance the CPU processing capability.We formulate these two problems as two optimization programs and prove their hard-ness.Based on parallel multi-block Alternating Direction Method of Multipliers(ADMM),we propose an(O(1),O(1))bicriteria approximation algorithm and a 2-approximation local search algorithm.Large-scale simulations and DPDK-based OpenNetVM platform show that Plutus can reduce the capital cost by 84%and increase the throughput by 36%on average.2.In the second part,we further propose a virtual network function deployment scheme such that multi-resource capacity constraints can be satisfied.This is because each network function may consume more than one type of resource when it is actually deployed on the server,which greatly increases the difficulty of deployment scheme design.We formulate Multi-resource VNF Deployment Problem(MVDP)as an optimization program and prove its hardness.We propose an offline(O(1),O(1))－bicriteria approximation algorithm and an(O(1),O(n·logn))-competitive online algorithm to deploy VNFs in a scalable manner,where n is the number of required VNFs for the arrived flows.Large-scale simulations and DPDK-based OpenNetVM implementation show that our algorithms for virtual network function deployment can reduce the overall cost by 34%on average and improve the performance in terms of multi-resource allocation.