首页|面向设备直通的高效低延时的中断直通方法

面向设备直通的高效低延时的中断直通方法

扫码查看
针对对称多处理器(SMP)虚拟机(VM)的虚拟中央处理器(vCPU)调度延迟会降低虚拟机输入/输出(I/O)响应性的问题,本文基于设备直通提出了一种高效低延迟的中断直通方法。该方法基于硬件辅助技术,搭建了中断直通架构,并设计了中断重定向机制,将直通设备中断从被抢占的vCPU重定向至正在运行的vCPU。实验结果表明,网络往返时延平均减少了 34。1%,吞吐量最高提升7。9%,Apache测试每个服务器请求所需时间平均减少了 13。6%,磁盘I/O操作时延平均减少了 6。7%~8。4%。实验结果证明,该方法能有效减少虚拟机虚拟CPU调度对I/O延迟的影响,提高虚拟机I/O响应性。
An efficient and low-latency interrupt passthrough method based on device passthrough
Aiming at the problem that the virtual central processing unit(CPU)scheduling delay of symmetric multi-pro-cessing(SMP)virtual machine(VM)will reduce the virtual machine input/output(I/O)responsiveness,an effi-cient and low-latency interrupt passthrough method based on device passthrough is proposed in this paper.Based on hardware-assisted technology,this method builds an interrupt pass-through architecture,and designs an interrupt redirection mechanism to redirect interrupts from the preempted vCPUs to running ones.The experimental results show that the round-trip time of the network is reduced by an average of 34.1%,the throughput is increased by up to 7.9%,and the time required for each server request with Apache test is reduced by an average of 13.6%,and the average latency of disk I/O operations is reduced by 6.7%~8.4%.The experimental results demonstrate that this method can effectively reduce the impact of virtual machine vCPU scheduling on I/O latency and improve virtu-al machine I/O responsiveness.

interrupt remappinginput/output(I/O)virtualizationdevice passthroughkernel-based virtu-al machine(KVM)I/O responsiveness

吕晨、张福新、朱琛、毛碧波、邓平科、潘筱涵

展开 >

处理器芯片全国重点实验室(中国科学院计算技术研究所)北京 100190

中国科学院计算技术研究所 北京 100190

中国科学院大学 北京 100049

龙芯中科技术股份有限公司 北京 100190

中国移动通信有限公司研究院 北京 100032

展开 >

中断重映射 输入/输出(I/O)虚拟化 设备直通 基于内核的虚拟机(KVM) I/O响应性

2024

高技术通讯
中国科学技术信息研究所

高技术通讯

CSTPCD北大核心
影响因子:0.19
ISSN:1002-0470
年,卷(期):2024.34(8)