摘要
在高端云服务器系统中,计算节点间的互连芯片通过Cache一致性协议将多计算节点互连组成分布式和共享内存空间系统,对接口传输速率和路由交换效率要求较高。文中通过分析Cache一致性协议报文的传输特点和互连网络转发需求,设计实现了一种互连芯片的高阶非对称交叉开关。设计通过了系统级的仿真验证,基于FPGA实现的云服务器互连芯片原型验证系统进行了实际带宽测试和芯片带宽匹配优化。互连芯片流片后的系统实测结果表明,满足功能要求,互连网络处理模块延迟8.75ns,吞吐率65.03%,达到了设计目标。
In the high-end cloud server system,the interconnect chip between computing nodes in-terconnect multiple computing nodes to form a distributed and shared memory space system by cache coherence protocol,therefore,high-speed interface communication and routing and forwarding play a vital role.The transmission characteristics of cache coherence protocol packets and the interconnection network forwarding requirements is analyzed in this paper,an asymmetric high-radix crossbar was realized.The design passed the system level simulation verification,based on a cloud server interconnect chip prototype verification system,the actual bandwidth test and bandwidth matching optimization was carried out.The actual system test after the chip tapout shows that the interconnect network processing module has a delay of 8.75ns and a throughput rate of 65.03%,which meets the functional requirements.
作者
邹晓峰
刘同强
周玉龙
李拓
李仁刚
公维锋
ZOU Xiao-feng;LIU Tong-qiang;ZHOU Yu-long;LI Tuo;LI Ren-gang;GONG Wei-feng(State Key Laboratory of High-end Server&Storage Technology,Jinan 250013,China;Inspur(Beijing)Electronic Information Industry Co.,Ltd.,Beijing 100085,China)
出处
《信息技术》
2019年第12期6-10,14,共6页
Information Technology
基金
北京市科委科技新星计划(2018077)
关键词
芯片互连
交叉开关
网络处理
路由转发
chip interconnect
crossbar
network processing
routing and forwarding