面向实时网络应用的虚拟网络功能部署

Virtual Network Function Deployment Facing Real-time Applications

相比较于传统的解决方案,网络功能虚拟化技术能够有效地减少网络应用的部署成本。实时网络应用是对于数据处理响应时间有着严格时间期限约束要求的一类应用,有着复杂的数据处理流程,通过调用底层的虚拟网络功能,可以完成对数据的实时处理。虚拟网络功能模块可以被部署到基质网络的不同节点中,这些节点具有异构性,同时具有不同的数据处理能力,此外,基质网络中节点间的延迟也有所不同。因此,如何有效地在基质网络中进行虚拟网络功能的部署,从而保证实时网络应用时间期限约束需求是当前亟需解决的问题。提出一种新型的DAP(Deadline-Aware VNF Placement)算法来部署实时应用的虚拟网络功能。该算法首先采用一种启发式的资源分配策略为每个实时应用分配优先级,然后按照优先级的大小为实时网络应用分配资源。采用TopologyZoo中的某些网络拓扑结构作为测试集,通过实验验证该方法可以有效地减少实时网络应用的时间期限丢失率。

Compared with traditional solutions,NFV(Network Function Virtualization)can help to reduce the cost of application deployment.Real-time network applications refer to the applications that have very strict time requirements.They have very complex data flows.These applications can complete the data processing by calling the basic VNFs(Virtual Network Function).These VNFs can be placed on different nodes in the substrate network.The nodes have different processing capabilities.The network latency between these nodes is also different.Thus,how to effectively place the VNFs in a substrate network and guarantee the time requirements of real-time applications has become an urgent problem.To solve this problem,this paper proposes DAP(Deadline-Aware VNF Placement)algorithm to place the VNFs of a real-time application.The algorithm first assigns priority through a heuristic method and allocates resources to the VNFs.This paper applies TopologyZoo as the underlying substrate network as dataset.Experimental results show that the proposed solution can effectively reduce the deadline missing rates of real-time network applications.