高级检索

    网络功能虚拟化中延时感知的资源调度优化方法

    Delay-Aware Resource Scheduling Optimization in Network Function Virtualization

    • 摘要: 网络功能虚拟化(network function virtualization, NFV)旨在以软件的方式实现网络功能从而替代传统网络中的专有硬件设备.为了应对日益增长的资源密集型需求,面向软件的网络功能虚拟化带来了如虚拟网络功能的管理、低延迟的调度和虚拟网络资源分配等问题.虚拟网络功能调度问题本身为NP-hard,在虚拟网络功能资源调度延迟的特定问题上,为保证良好的用户体验,需要确保网络资源被合理地分配和协调以防止资源的过度供应和保持端到端低延迟.针对网络资源调度的延时问题建立了以最小化资源调度总体服务延迟为目标的整数线性规划模型.此外,为了满足网络动态性较高的特性,设计了一种基于贪婪的启发式算法,此算法首先构建辅助图,然后根据考虑到网络传播时延影响的不同业务链之间的时延影响分析来选择资源调度方案,并且对很多点处理功能采用了多路传输的方式.最终的实验结果表明:所提算法可以有效地指导模型的求解,在降低网络总体服务延时方面比之前相关研究有5%~15%的性能提升.

       

      Abstract: Network function virtualization (NFV) aims to implement software-enable network functions so as to replace proprietary hardware devices in traditional networks. In response to the growing need for intensive resource, the software-oriented network functions bring issues such as the management of virtual network functions, low-latency scheduling, and efficient allocation of virtual network resources. The virtual network function scheduling problem itself is NP-hard. In order to ensure a good user experience on the specific issues of scheduling delays for virtual network function resources, it is necessary to ensure that network resources are reasonably allocated to prevent over-provisioning of resources and to maintain low end-to-end latency. In this paper,we formulate the problem as an integer linear programming model. In addition, to meet the characteristics of high dynamic network, we design a heuristic algorithm based on greedy. This algorithm firstly constructs the auxiliary graph, and then selects the node resources according to the analysis of the delay impact among different service chains, which considers the influence of network propagation delay and adopts multi-path transmission scheme for the multipoint processing function. The final experimental results show that our algorithm can effectively solve the model and reduce the overall network service latency by 5% to 15% compared with the previous research.

       

    /

    返回文章
    返回