面向数据中心的软件定义网络技术研究

发布时间：2018-11-22 10:48

【摘要】：作为云计算的核心基础设施,数据中心网络是一个连接了数万级、十万级甚至百万级的大规模服务器群来进行大型分布式计算的桥梁,因此更成为了互联网研究热点中的热点。随着网络规模的膨胀,数据中心所需处理的数据信息也随之呈现出爆炸式的增长趋势,建设管理复杂、控制不灵活等问题也愈加明显,进而影响了数据中心网络的进一步发展。而通过利用软件定义网络(Software Defined Networking,SDN)技术来搭建数据中心,不仅可以简化数据中心的网络管理,而且还能优化其网络性能,进一步提高数据中心的业务承载能力,因而受到了工业界和学术界的广泛重视。本文首先对数据中心网络、软件定义网络以及当前数据中心网络的SDN技术应用方案进行了细致的研究,并重点研究了 SDN实现的关键技术Openflow及其网络数据报文的转发机制,同时分析了现有基于TCAM(Ternary Content Addressable Memory,三态内容寻址存储器)实现的Openflow流表转发策略,在研究过程中我们发现Openflow在实际的应用中仍然存在着一些技术瓶颈,例如:路由器/交换机中广为使用的快速查找TCAM存储器成本的问题以及内存问题。为此本文提出了 CMDDBF(Combined Multi-Dimension Dynamic Bloom Filter,联合多维动态Bloom过滤器)查找结构来替代Openflow交换机中的TCAM实现流表的转发和处理。CMDDBF解决了 TCAM空间利用率低和价格昂贵的缺点,能够有效实现多字段数据的查找和处理操作,并根据流表的规模动态的调整内存空间。CMDDBF具有良好的灵活性和可扩展性,更适用于Openflow交换机流表的查找和处理操作。通过实验分析,CMDDBF在降低误识率方面具有很好的性能。。
[Abstract]:As the core infrastructure of cloud computing, data center network is a bridge that connects tens of thousands, 100,000 or even millions of large-scale server clusters to carry out large-scale distributed computing, so it has become a hot spot in Internet research. With the expansion of the network scale, the data information that the data center needs to deal with also presents the explosive growth trend, the construction management is complex, the control is inflexible and so on, has affected the further development of the data center network. By using software definition network (Software Defined Networking,SDN) technology to build data center, not only the network management of data center can be simplified, but also its network performance can be optimized, and the service carrying capacity of data center can be further improved. As a result, the industry and academia have been widely valued. In this paper, the application scheme of SDN technology in data center network, software definition network and current data center network is studied in detail, and the key technology of SDN realization, Openflow and the forwarding mechanism of network data message, are studied in detail in this paper. At the same time, the Openflow stream table forwarding strategy based on TCAM (Ternary Content Addressable Memory, three-state content addressable memory is analyzed. In the process of research, we find that there are still some technical bottlenecks in the practical application of Openflow. For example, the cost of fast lookup of TCAM memory and memory problem in routers / switches. In this paper, CMDDBF (Combined Multi-Dimension Dynamic Bloom Filter, combined with multidimensional dynamic Bloom filter is proposed to replace TCAM in Openflow switch to realize stream table forwarding and processing. CMDDBF solves the disadvantages of low TCAM space utilization and high price. CMDDBF has good flexibility and extensibility, and it is more suitable for the search and processing of flow table in Openflow switch. Through experimental analysis, CMDDBF has good performance in reducing error rate.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP308

【参考文献】