时间敏感网络延时组成的精确分析与测量方法
2024,46(5):189-199
付文文
国防科技大学 计算机学院, 湖南 长沙 410073，fuwenwen94@nudt.edu.cn,sunzhigang@nudt.edu.cn
全巍
国防科技大学 计算机学院, 湖南 长沙 410073
姜旭艳
国防科技大学 计算机学院, 湖南 长沙 410073
孙寅涵
国防科技大学 计算机学院, 湖南 长沙 410073
孙志刚
国防科技大学 计算机学院, 湖南 长沙 410073，fuwenwen94@nudt.edu.cn,sunzhigang@nudt.edu.cn
国防科技大学 计算机学院, 湖南 长沙 410073，fuwenwen94@nudt.edu.cn,sunzhigang@nudt.edu.cn
全巍
国防科技大学 计算机学院, 湖南 长沙 410073
姜旭艳
国防科技大学 计算机学院, 湖南 长沙 410073
孙寅涵
国防科技大学 计算机学院, 湖南 长沙 410073
孙志刚
国防科技大学 计算机学院, 湖南 长沙 410073，fuwenwen94@nudt.edu.cn,sunzhigang@nudt.edu.cn
摘要：
在依据时间敏感网络(time-sensitive networking, TSN)标准估算相邻节点之间的最大传输延迟Δt时,将不可避免地引入额外的无效等待时间,这种等待时间被称为“泡沫延时”。分析了泡沫延时对增加端到端延时和降低规划成功率的负面影响,并通过细粒度地分析Δt的延时组成,首次提出Δt的精确测量方法。基于精确的Δt,消除了TSN规划时产生的泡沫延时。基于两款定制的TSN交换设备搭建了真实的测试环境,测试结果显示,泡沫延时至少占端到端延时的26.4%,并且消除泡沫延时后规划成功率提升了8.9%~39.1%。
在依据时间敏感网络(time-sensitive networking, TSN)标准估算相邻节点之间的最大传输延迟Δt时,将不可避免地引入额外的无效等待时间,这种等待时间被称为“泡沫延时”。分析了泡沫延时对增加端到端延时和降低规划成功率的负面影响,并通过细粒度地分析Δt的延时组成,首次提出Δt的精确测量方法。基于精确的Δt,消除了TSN规划时产生的泡沫延时。基于两款定制的TSN交换设备搭建了真实的测试环境,测试结果显示,泡沫延时至少占端到端延时的26.4%,并且消除泡沫延时后规划成功率提升了8.9%~39.1%。
基金项目：
国防科技大学并行与分布处理国家级实验室开放基金资助项目(WDZC20205500110)
国防科技大学并行与分布处理国家级实验室开放基金资助项目(WDZC20205500110)
Precise analysis and measurement method for delay composition of time-sensitive networking
FU Wenwen
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China,fuwenwen94@nudt.edu.cn,sunzhigang@nudt.edu.cn
QUAN Wei
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China
JIANG Xuyan
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China
SUN Yinhan
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China
SUN Zhigang
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China,fuwenwen94@nudt.edu.cn,sunzhigang@nudt.edu.cn
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China,fuwenwen94@nudt.edu.cn,sunzhigang@nudt.edu.cn
QUAN Wei
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China
JIANG Xuyan
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China
SUN Yinhan
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China
SUN Zhigang
College of Computer Science and Technology, National University of Defense Technology, Changsha 410073, China,fuwenwen94@nudt.edu.cn,sunzhigang@nudt.edu.cn
Abstract：
When estimating the maximum transmission delay Δt between adjacent nodes according to TSN (time-sensitive networking) standards, it was inevitable to introduce additional ineffective waiting time, which was referred to as “bubble delay”. The negative effect of bubble delay on increasing the end-to-end delay and decreasing the planning success rate was analyzed, and the precise measurement method of Δt was proposed for the first time based on analyzing the delay composition of Δt in fine granularity. Based on the exact Δt, the bubble delay caused by TSN planning was eliminated. A real test environment was built based on two customized TSN switching devices. The test results show that the bubble delay accounted for 26.4% of the end-to-end delay at least, and the planning success rate increased by 8.9% to 39.1% after eliminating the bubble delay.
When estimating the maximum transmission delay Δt between adjacent nodes according to TSN (time-sensitive networking) standards, it was inevitable to introduce additional ineffective waiting time, which was referred to as “bubble delay”. The negative effect of bubble delay on increasing the end-to-end delay and decreasing the planning success rate was analyzed, and the precise measurement method of Δt was proposed for the first time based on analyzing the delay composition of Δt in fine granularity. Based on the exact Δt, the bubble delay caused by TSN planning was eliminated. A real test environment was built based on two customized TSN switching devices. The test results show that the bubble delay accounted for 26.4% of the end-to-end delay at least, and the planning success rate increased by 8.9% to 39.1% after eliminating the bubble delay.
收稿日期：
2022-05-12
2022-05-12
