软件定义网络(Software Defined Networking，SDN)是一种革命性的网络架构，主要思想是将控制平面与数据平面分离，并且还拥有开放可编程特性。其对数据包转发以及网络资源管理方面有着极高的要求。三态内容寻址存储器(Ternary Content Addressable Memory，TCAM)因其快速规则匹配能力通常作为规则的缓存应用于SDN交换机中。规则缓存将大部分流量引导到高性能的硬件路径上，可以显著提升网络性能。然而，规则之间存在的依赖关系使得缓存的利用率变低。因此，合理的规则缓存算法对本就稀有的 TCAM 资源来说十分重要。聚焦规则间的依赖关系，该文提出了CacheBand规则缓存算法。该算法利用包围盒思想，通过对规则及当前流量的分析，智能产生绷带规则，切断了规则间的依赖关系。实验证明，在不同数据包速率下，与同类算法相比，CacheBand可减少约68%的缓存条目，显著降低了流表压力，为数据转发提供了可靠的缓存方案。
A Bounding Box Based Efficient Storage SDN Rule Caching Algorithm
Software Defined Networking(SDN)has revolutionized network architecture by decoupling the control plane from the data plane and introducing open programmable features.This innovation imposes rigorous demands on tasks like packet forwarding and network resource management.Ternary Content Addressable Memory(TCAM)is a crucial cache component in SDN switches due to its swift rule matching capability.Rule caching optimizes traffic routing,guiding it towards high-performance hardware paths,resulting in a substantial enhancement of overall network efficiency.However,the intricacies of rule dependencies pose a challenge,leading to suboptimal cache utilization.This underscores the critical need for a sophisticated rule caching algorithm,especially given the scarcity of TCAM resources.Inspired by the bounding box concept,we introduce the CacheBand rule caching algorithm.This algorithm,after a me-ticulous analysis of rules and current traffic,dynamically generates bandage rules.These bandage rules can dynamically adjust to real-time network changes,ensuring the stability of the caching algorithm while cutting dependencies between rules.Empirical validation of CacheBand demonstrates a remarkable reduction of approximately 68%in cache entries across varying packet rates,outperforming similar algorithms.This achievement substantially eases the burden on flow tables,providing a robust and dependable caching solution for seamless network data forwarding.
software defined networkingternary content addressable memoryrule dependencyrule cachingbounding box
国家科技期刊平台
NSTL
万方数据
