面向软件定义网络的流表优化方案

摘要

针对目前软件定义网络中细粒度的流匹配机制造成的网络流表项空间开销和查询开销爆炸式增长等问题,提出了一种全新的基于布隆过滤器(Bloom Filter)的多级流表结构.该结构为混合流表结构,采用Bloom Filter多级流表结构来存储流表项,主要着眼于提高软件定义网络(SDN)交换机流表的容量和加快流表项的匹配速度;在流表项语义层面,设计并实现了控制器与SDN交换机之间的中间适配层模块来解决语义冲突问题.基于真实流量的实验结果表明,在规则占用空间上,与传统流表相比,Bloom Filter在流表越精细的情况下优化比率越高,最高可达90.7％.随着流表项规则的增加,匹配耗时优化效率提高,匹配时间最多可减少99.4％.该问题的解决可望为SDN网络的大规模实用化部署奠定数据层面的基础.%A new multilevel flow table architecture based on Bloom Filter is proposed to solve the problem of the explosive growth of the flow table's space occupation and look-up overhead due to the fine grained flow matching mechanism in SDN networks.This architecture follows the hybrid flow table paradigm and uses a Bloom Filter pipeline to store flow entries.The motivation is to construct secondary flow tables with very high capacity and to speed up the matching speed of flow table items.An intermediate adaptation layer between controllers and OpenFlow switches is designed to transform the corresponding messages and to avoid possible semantic conflicts.Experiments results show that when the flow table becomes more fine-grained,the space occupation of the proposed scheme is more effectively reduced,up to 90.7％,and when the number of the flow entries increases,the packet matching time is greatly reduced and the maximum reduction of matching time is 99.4％.The solution of the problem is expected to lay a solid foundation in data plane for the large-scale practical deployment of SDN networks.