检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
作 者:王继昌 吕高锋[1] 刘忠沛 杨翔瑞 WANG Jichang;LU Gaofeng;LIU Zhongpei;YANG Xiangrui(School of Computer,National University of Defense Technology,Changsha 410073,China)
出 处:《小型微型计算机系统》2024年第3期692-698,共7页Journal of Chinese Computer Systems
基 金:国家重点研发计划项目(2020YFB1805603)资助.
摘 要:随着网络规模的进一步扩展,传统B/S架构的单体应用逐渐被微服务所代替,服务的拆分使得API的规模呈指数增长.而商用网卡处理在处理海量的微服务请求时,在确定性、可编程和数据拷贝三个方面表现出巨大的局限性.为保证各网络节点高效、灵活、精确地处理服务请求,本文提出可编程确定性的多队列FPGA加速器原型.该加速器依托多个硬件队列以及队列管理单元,扩展基于规则的RSS算法,实现serverless友好的数据包分发.添加PTP硬件时钟组件,与队列管理单元协同控制对数据包的确定性发送.为提升网络节点的数据收发效率,设计了适配FPGA加速器的驱动程序,实现数据包收发的零拷贝.在支持100Gbps线速率的FPGA上进行的实验表明,该加速器可以支持不同数据包大小的多队列的零拷贝传输,定义网路节点的转发行为,并在8个核心的FPGA设备上接近线速率处理数据,此外,它还支持以接近ovs-DPDK的低延迟进行数据包转发,在一定程度上克服了商业网卡在面向微服务架构时的局限性.With the expansion of network scale,the traditional B/S architecture of monolithic applications is gradually replaced by microservices.The splitting of services makes the size of APIs grow at the fingertips.The commercial NIC processing shows huge limitations in three aspects:deterministic,programmable,and data copy when handling massive microservice requests.To ensure that each node handles service requests efficiently,flexibly,and precisely,this paper proposes a programmable deterministic multi-queue FPGA gas pedal prototype.The accelerator relies on the 1000 hardware queues and the queue management unit to extend the rule-based RSS algorithm for serverless-friendly programmability of packet distribution.A PTP hardware clock component is added to collaborate with the queue management unit to control the deterministic delivery of packets.To improve network node data sending and receiving efficiency,a driver adapted to the FPGA gas pedal is designed to achieve zero-copy of packet sending and receiving.Experiments conducted on a 100Gbps FPGA show that the card can support the multi-queue transmission with different packet sizes,define the forwarding behaviour of network nodes,and approach the line rate on an 8-core FPGA device.In addition,it can forward packets with low latency close to ovs-DPDK,which overcomes to some extent the commercial NICs when oriented to microservice architecture Limitations.
分 类 号:TP393[自动化与计算机技术—计算机应用技术]
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:18.219.203.214