一种求解服务链映射问题的离散粒子群优化算法  

Discrete Particle Swarm Optimization Algorithm for Service Chain Mapping Problem

作　　者：马丁[1,2] 庄雷[1] 兰巨龙[3] 

机构地区：[1]郑州大学信息工程学院,郑州450001 [2]河南工业大学信息科学与工程学院,郑州450001 [3]国家数字交换系统工程技术研究中心,郑州450002

出　　处：《小型微型计算机系统》2017年第8期1811-1817,共7页Journal of Chinese Computer Systems

基　　金：国家"九七三"重点基础研究发展计划基金项目(2012CB315901)资助;国家自然科学基金项目(61379079)资助;河南省国际合作项目(152102410021)资助

摘　　要：网络功能虚拟化技术的提出,为减少服务提供商的投资成本与管理开销,实现网络功能的灵活组合、新服务的快速部署提供了解决方案,使多样化、可扩展的服务供应成为可能.网络功能虚拟化中的研究难点是服务链映射问题.现有以最小化端到端时延为目标的映射算法受制于资源瓶颈的限制,导致有效服务路径不能被映射,影响服务链的映射成功率.针对这一问题,提出一种求解服务链映射问题的粒子群优化算法(SIFC-SCMPSO),并以路径跳数为约束设计了初始化可行解与位置更新策略.实验结果表明,该算法有效地减少了资源用量,提高了有效路径被发现的概率,在服务链映射成功率、长期平均收益、平均开销等评价指标上均取得了较好的结果.The rise of Network Function Virtualization （NFV） introduces solutions for Interact Service Providers （ISPs） to reduce capital expenditures and operational cost, and for network operators to flexibly composite network functions and swiftly deploy novel services. NFV enables diverse and scalable service provisioning. A major challenge in NFV is how to map service chains onto the physical network. Existing mapping algorithms taking minimizing end-to-end latency into account are incapable of dealing with resource bottlenecks,which leads to low mapping success rate since the valid path cannot be found even when a path exists. To address the issue, we propose a discrete particle swarm optimization algorithm for service chain mapping problem （ SIFC-SCMPSO ） and design a solution initialization and update strategy considering the hop constraint of physical paths. Our experiments suggest that the proposed algorithm decreases the resource usage efficiently,increase the probability of finding the valid path,and improves the mapping performance significantly in terms of service chain mapping success rate and long-term average revenue while decreasing the average cost.

关 键 词：网络功能虚拟化 服务链 映射 服务路径 粒子群优化 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]