论文部分内容阅读
在未来互联网研究领域中,虚拟化路由器受到越来越多的关注。作为实现网络虚拟化的关键路由设备,虚拟化路由器在一个物理路由平台基础上拥有多个虚拟路由实例,每一个路由实例维护自己的转发表。因此,有限的片上存储限制了存储的扩展性。本文中,我们提出一种基于融合再拆分的方法,用于压缩多个转发表,不仅提高了存储效率,同时为快速系统重构过程提供了一个理想的拆分位置。另外,本文提出了一种优化策略,用于减少快速系统重构的时间。实验表明,我们的方案在处理14个真实路由数据集时,只需要一棵134 188个结点的单步长特里树,而原始方案中需要4 569 133个结点。同时,我们的方案在扩展性中表现出良好的性能,能够支持90 000 000条前缀以及65 600个转发表。
In the future of Internet research, virtualized routers are receiving more and more attention. As a key routing device for network virtualization, a virtual router has multiple virtual routing instances based on one physical routing platform, and each routing instance maintains its own forwarding table. Therefore, limited on-chip storage limits storage scalability. In this paper, we propose a fusion-based re-splitting method for compressing multiple forwarding tables, which not only improves the storage efficiency, but also provides an ideal splitting position for rapid system reconfiguration. In addition, this paper presents an optimization strategy to reduce the time for rapid system reconfiguration. Experiments show that when we deal with 14 real routing datasets, we only need one single-stepping terrestrial tree of 134 188 nodes, compared with 4 569 133 nodes in the original scheme. At the same time, our solution shows good performance in scalability, capable of supporting 90,000,000 prefixes and 65,600 forwarding tables.