Network Function Virtualization: Placement and Performance Optimization

网络功能虚拟化（NFV）和网络服务链（SC）范式改变了网络功能和服务在底层网络基础设施中的部署方式。现在，网络功能与专用的中间件解耦，可以作为软件实体提供，称为虚拟网络功能（VNFs），可以在行业标准服务器中实例化，而完整的网络服务可以通过部署一系列VNFs创建服务链模型来提供。这种部署模型使得服务提供商（sp）能够根据最终用户的需求提供高度专业化的服务，同时降低了总体资本和运营费用。然而，总成本的降低通常是实现吞吐量性能的函数。因此，本提案的目标是利用行业标准服务器的功能，并在托管VNFs时增强其性能，以满足专用硬件设备的高效率，并最大限度地降低总体部署成本。解决这个问题的好处如下：a)通过使用更少的硬件资源最大限度地降低功耗来促进“绿色”架构；b)通过在同一台服务器上安装多种网络功能来降低部署成本；c)加快上市速度，最大限度地缩短网络运营商提供新服务的创新周期；d)将NFV与云计算和软件定义网络（SDN）技术相结合，为多个用户提供网络服务，而不考虑其地理分布；e)根据用户的需求，通过适当地扩大/缩小服务，实时优化网络配置。为此，将提出适当的分配算法，将服务链的VNFs优化部署在标准的现成服务器池中，以最小化总体部署成本，最大化实现吞吐量性能。

Network Functions Virtualization (NFV) along with the Network Service Chaining (SC) paradigm havechanged the way that network functions and services are deployed in the underlying network infrastructure.Network functions are now decoupled from dedicated middleboxes and can be offered as software entitiescalled Virtualized Network Functions (VNFs) that can be instantiated in industry standard servers, whilecomplete network services can be provided by deploying a series of VNFs creating a Service Chain model.This deployment model has led the Service Providers (SPs) to offer highly specialized services tailored to theneeds of end users reducing at the same time the overall capital and operational expenses. Total cost reductionshowever are typically a function of the throughput performance achieved. Hence, the goal of this proposal is toleverage the capabilities of industry standard servers and enhance their performance when hosting VNFs tomeet the high efficacy of the dedicated hardware appliances and minimize the overall deployment cost. Thebenefits of addressing this problem are the following: a) promote a "green" architecture by minimizing powerconsumption using less hardware resources, b) reduce deployment cost by installing multiple network functionson the same server c) increase the velocity of Time to Market minimizing the typical network operator cycle ofinnovation to offer new services, d) provide network services to multiple users regardless their geographicallocation by using NFV in conjunction with the Cloud Computing and Software Defined Networking (SDN)technologies e) optimize the network configuration in real time by appropriately scaling up/down the servicesaccording to the users' requirements. To this end, appropriate allocation algorithms will be proposed to optimaldeploy the VNFs of a Service Chain in a pool of standard off the shelf servers, with the goal to minimize theoverall deployment cost while maximizing the achieved throughput performance.