• 一种基于共享缓存的星载数据交换架构设计

    Design of a shared-memory-based spaceborne data switching architecture

    • 针对低轨卫星数据交换系统对吞吐量提升与突发流量吸收的迫切需求,提出一种两级仲裁–分区缓存交换架构。该架构采用分级仲裁策略,实现多端口并行访问优化,有效缓解了资源共享冲突导致的吞吐量瓶颈问题。系统集成动态内存管理机制,支持16个千兆以太网端口全双工存储转发操作,并内嵌具备2位检错/1位纠错能力的ECC校验单元及多优先级QoS调度算法,以增强数据传输的可靠性与服务质量保障。在Xilinx XC7V690T FPGA平台上完成RTL级设计、综合实现及系统测试。实验结果表明,在156.25 MHz工作频率下,系统平均端口线速达4 Gbps,总吞吐量达到64 Gbps,时序裕度保持在20%以上。该架构在吞吐性能、时序收敛与硬件资源开销之间实现了良好平衡,为星载交换系统的设计与优化提供了有效参考。

       

      Abstract: To address the critical demands for throughput enhancement and burst traffic absorption in low-earth-orbit (LEO) satellite data switching systems, this paper proposes a two-level arbitration with partitioned cache switching architecture. The architecture employs a hierarchical arbitration strategy to optimize multi-port parallel access, effectively alleviating the throughput bottleneck caused by resource-sharing conflicts. The system integrates a dynamic memory management mechanism, supports full-duplex store-and-forward operations across 16 Gigabit Ethernet ports, and incorporates an ECC checking unit with 2-bit error detection and 1-bit error correction capability, along with a multi-priority QoS scheduling algorithm to enhance data transmission reliability and service quality assurance. RTL-level design, synthesis, implementation, and system testing are completed on the Xilinx XC7V690T FPGA platform. Experimental results demonstrate that at an operating frequency of 156.25 MHz, the system achieves an average port line rate of 4 Gbps, with a total system throughput of 64 Gbps and a timing margin exceeding 20%. The proposed architecture achieves a favorable balance among throughput performance, timing closure, and hardware resource overhead, providing a valuable reference for the design and optimization of spaceborne switching systems.

       

    /

    返回文章
    返回