基于SDRAM的精确查找方法研究与设计
发布时间：2019-05-17 23:19
【摘要】：随着网络技术的高速发展,互联网已经成为人们生活中不可或缺的部分。网络业务流量呈每六个月倍增的趋势。随着光通信技术的发展,链路传输速率已不再是网络技术的性能瓶颈,而网络查找转发处理则成了制约互联网发展的关键因素之一。针对现有网络管理不灵活、流量调度不均匀的问题,软件定义网络(Soft-Defined Network,SDN)的概念被提出来。在SDN的核心协议OpenFlow中,查找以流表的形式存在。本文利用片外存储器设计查找表,重点研究了适合OpenFlow交换机的精确查找方法。 本论文首先对现有的高速查找方法进行了调研和分析。结合基于前缀的路由查找方法和OpenFlow交换机中流表的形式,提出了一种基于SDRAM(Synchronous Dynamic Random Access Memory)的精确查找方法。该方法主要包括硬件预处理和片外查找表两个方面。在硬件预处理方面,提出了基于并行布隆过滤器的硬件预处理方法,用于减少一次成功查找所需要片外存储器的访问次数。经过片内资源的合理分配和布隆过滤器的改进,最终一次成功查找所需要的片外访问的次数近似为一次。然后,结合OpenFlow中流表需要具备可扩展性和并行查找的需求,论文提出了一种容量可配置的并行布隆过滤器,最后为了降低片内处理的功耗,提出了一种两级流水线型的布隆过滤器,当配置合适时,最多可以节省80%的功耗。在片外查找表方面,本文研究了片外存储器SDRAM指令重排机制,以提高片外缓存的总线利用率。本文提出的基于突发交错的指令重排机制与现有的突发调度重排机制相比总线利用率提高了25%。然后本文根据OpenFlow流表的形式和SDRAM器件特性设计了适合流表查找的SDRAM地址分配方案。最后本文研究了SDRAM的控制原理和SDRAM控制器的设计方法,并利用FPGA实现了SDRAM控制器 OpenFlow交换机中流表的查找形式有利于更灵活地控制网络,但是多流表查找也对高速查找算法提出了更高的要求,所以对适用于OpenFLow流表查找算法的研究具有重要的意义。
[Abstract]:With the rapid development of network technology, the Internet has become an indispensable part of people's lives. Network traffic shows a trend of doubling every six months. With the development of optical communication technology, link transmission rate is no longer the performance bottleneck of network technology, and network search and forwarding processing has become one of the key factors restricting the development of the Internet. In order to solve the problems of inflexible network management and uneven traffic scheduling, the concept of software defined network (Soft-Defined Network,SDN) is put forward. In OpenFlow, the core protocol of SDN, search exists in the form of flow table. In this paper, the off-chip memory is used to design the lookup table, and the accurate search method suitable for OpenFlow switch is studied. In this paper, the existing high-speed search methods are investigated and analyzed. Based on the prefix based routing lookup method and the form of flow table in OpenFlow switch, an accurate search method based on SDRAM (Synchronous Dynamic Random Access Memory) is proposed. This method mainly includes two aspects: hardware preprocessing and off-chip look-up table. In the aspect of hardware preprocessing, a hardware preprocessing method based on parallel Bron filter is proposed to reduce the number of visits to the off-chip memory needed for a successful search. After the reasonable allocation of on-chip resources and the improvement of Bron filter, the number of off-chip visits required for a successful search is approximately one. Then, in order to reduce the power consumption of on-chip processing, a parallel Bron filter with configurable capacity is proposed to meet the needs of scalability and parallel search of flow table in OpenFlow. A two-stage streamline Bron filter is proposed, which can save up to 80% power consumption when it is properly configured. In the aspect of off-chip look-up table, this paper studies the SDRAM instruction rearrangement mechanism of off-chip memory in order to improve the bus utilization of off-chip cache. Compared with the existing burst scheduling rearrangement mechanism, the bus utilization of the instruction rearrangement mechanism based on burst interleaving proposed in this paper is 25% higher than that of the existing burst scheduling rearrangement mechanism. Then, according to the form of OpenFlow flow table and the characteristics of SDRAM device, a SDRAM address assignment scheme suitable for stream table search is designed in this paper. Finally, this paper studies the control principle of SDRAM and the design method of SDRAM controller, and uses FPGA to realize the look-up form of flow table in SDRAM controller OpenFlow switch, which is beneficial to more flexible control network. However, multi-stream table search also puts forward higher requirements for high-speed search algorithm, so it is of great significance to study the algorithm suitable for OpenFLow flow table search.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.07
本文编号：2479485
[Abstract]:With the rapid development of network technology, the Internet has become an indispensable part of people's lives. Network traffic shows a trend of doubling every six months. With the development of optical communication technology, link transmission rate is no longer the performance bottleneck of network technology, and network search and forwarding processing has become one of the key factors restricting the development of the Internet. In order to solve the problems of inflexible network management and uneven traffic scheduling, the concept of software defined network (Soft-Defined Network,SDN) is put forward. In OpenFlow, the core protocol of SDN, search exists in the form of flow table. In this paper, the off-chip memory is used to design the lookup table, and the accurate search method suitable for OpenFlow switch is studied. In this paper, the existing high-speed search methods are investigated and analyzed. Based on the prefix based routing lookup method and the form of flow table in OpenFlow switch, an accurate search method based on SDRAM (Synchronous Dynamic Random Access Memory) is proposed. This method mainly includes two aspects: hardware preprocessing and off-chip look-up table. In the aspect of hardware preprocessing, a hardware preprocessing method based on parallel Bron filter is proposed to reduce the number of visits to the off-chip memory needed for a successful search. After the reasonable allocation of on-chip resources and the improvement of Bron filter, the number of off-chip visits required for a successful search is approximately one. Then, in order to reduce the power consumption of on-chip processing, a parallel Bron filter with configurable capacity is proposed to meet the needs of scalability and parallel search of flow table in OpenFlow. A two-stage streamline Bron filter is proposed, which can save up to 80% power consumption when it is properly configured. In the aspect of off-chip look-up table, this paper studies the SDRAM instruction rearrangement mechanism of off-chip memory in order to improve the bus utilization of off-chip cache. Compared with the existing burst scheduling rearrangement mechanism, the bus utilization of the instruction rearrangement mechanism based on burst interleaving proposed in this paper is 25% higher than that of the existing burst scheduling rearrangement mechanism. Then, according to the form of OpenFlow flow table and the characteristics of SDRAM device, a SDRAM address assignment scheme suitable for stream table search is designed in this paper. Finally, this paper studies the control principle of SDRAM and the design method of SDRAM controller, and uses FPGA to realize the look-up form of flow table in SDRAM controller OpenFlow switch, which is beneficial to more flexible control network. However, multi-stream table search also puts forward higher requirements for high-speed search algorithm, so it is of great significance to study the algorithm suitable for OpenFLow flow table search.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.07
【参考文献】
相关期刊论文 前3条
1 李振强;郑东去;马严;;TSB:一种多阶段IPv6路由表查找算法[J];电子学报;2007年10期
2 胥小波;郑康锋;李丹;杨义先;钮心忻;;基于并行BP神经网络的路由查找算法[J];通信学报;2012年02期
3 崔宇;田志宏;张宏莉;方滨兴;;基于前缀区间集合的IPv6路由查找算法[J];通信学报;2013年06期
相关博士学位论文 前1条
1 孙琼;下一代互联网的报文标识与查找技术的研究[D];北京邮电大学;2010年
，本文编号：2479485
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2479485.html
