Comp-TCAM: An Adaptable Composite Ternary Content-Addressable Memory on FPGAs
Comp-TCAM: An Adaptable Composite Ternary Content-Addressable Memory on FPGAs
复制标题
Comp-TCAM：FPGA 上的适应性复合三态内容可寻址存储器
DOI：
--
复制
发表时间：
2022
影响因子：
1.6
通讯作者：
R. Cheung
中科院分区：
文献类型：
--
作者：
Muhammad Irfan;Hasan Erdem Yantır;Z. Ullah;R. Cheung
采用具有类似软件的可重构性和类似硬件的性能的现场可编程门阵列（FPGA）作为开发平台来实现复杂系统，即，软件定义网络（SDN）。三进制内容可寻址存储器（TCAM）在现代FPGA中不存在，而是在需要时用作软核。FPGA的硬件结构是灵活但刚性的;基本结构是固定的，可以以有限的方式使用。现有的基于FPGA的TCAM耗尽一个特定类型的存储器时，一个大的尺寸是实现，导致系统的其余部分的内存短缺。我们提出的架构只使用那些存储器元件的FPGA是冗余的，不使用其他部分。在这封信中，所提出的架构，comp-TCAM，结合块RAM（BRAM）和查找表RAM（LUTRAM）来实现TCAM架构，消除了对存储器类型的依赖，并适应系统的要求。评估结果表明，可行性，可扩展性和有效性的建议TCAM架构相比，现有的TCAM架构。与最先进的基于FPGA的TCAM相比，Xilinx VIrtex-7 FPGA的硬件资源利用率降低了41.6%，而不会损害系统性能，提供21.06 Gbits/s的吞吐量。
Field-programmable gate arrays (FPGAs) having software-like reconfigurability and hardware-like performance are adopted as developing platforms to implement complex systems, i.e., software-defined networks (SDNs). Ternary content-addressable memory (TCAM) is not present in modern FPGA, but rather it is used as a softcore where needed. The hardware structure of an FPGA is flexible but rigid; the elemental structure is fixed and can be used in a limited number of ways. Existing FPGA-based TCAMs exhaust one particular type of memory when a large size is implemented, resulting in a shortage of memory for the rest of the system. Our proposed architecture uses only those memory elements of FPGA that are redundant and not used by other parts. In this letter, the proposed architecture, comp-TCAM, combines both block RAM (BRAM) and lookup table RAM (LUTRAM) to implement the TCAM architecture; that eliminates the dependency on the type of memory and is adaptable to the requirement of the system. Evaluation results show the feasibility, scalability, and effectiveness of the proposed TCAM architecture compared to the existing TCAM architectures. The hardware resource utilization on Xilinx VIrtex-7 FPGA is reduced by 41.6% compared to state-of-the-art FPGA-based TCAM with no harm to the system’s performance providing a throughput of 21.06 Gbits/s.