Status-Byte-Assisted RDMA Transmission Mechanism for Optimizing Multi-Task Video Streaming in Edge Computing

Author:

Xiao Donglei1,Yi Huiyue2ORCID,Zhang Wuxiong2ORCID,Shen Wenhui1ORCID

Affiliation:

1. School of Communication and Information Engineering, Shanghai University, Shanghai 200444, China

2. Science and Technology on Microsystem Laboratory, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 201899, China

Abstract

In the context of the rapid development of edge computing, optimizing data transmission and reducing latency is crucial for efficient collaborative processing among edge servers. Traditional TCP/IP protocols are hindered by high latency and low throughput, while RDMA (Remote Direct Memory Access) technology addresses these challenges by enabling direct memory access and bypassing the operating system kernel. However, the RDMA data transmission mechanism based on sliding windows requires frequent memory status exchanges in the order of memory blocks, which can limit its ability to handle multiple concurrent tasks within a single Queue Pair (QP). To address the limitations of the traditional sliding window transmission mechanism in multi-task environments, we propose a novel RDMA data transmission mechanism that utilizes status bytes to indicate memory block utilization, which utilizes stateless server connections, and multi-task shared QP transmission strategies. In the proposed mechanism, fine-grained control over memory blocks is achieved through the status byte, thereby enabling effective multi-task real-time video stream transmission. Experimental results show that, compared to the sliding window method, the proposed status-byte-assisted RDMA transmission mechanism provides higher throughput, lower latency, and reduced resource consumption, thus enhancing system scalability and reducing CPU utilization. Moreover, this mechanism achieves more stable throughput than the sliding window method when transmitting multiple real-time video streams in edge computing scenarios, making it particularly suitable for data transmission in such environments.

Funder

Special Projects for Key R&D Tasks in the Autonomous Region of Xinjiang

Publisher

MDPI AG

Reference32 articles.

1. Edge computing: Vision and challenges;Shi;IEEE Internet Things J.,2016

2. Zhou, X., Ke, R., Yang, H., and Liu, C. (2021). When intelligent transportation systems sensing meets edge computing: Vision and challenges. Appl. Sci., 11.

3. Liu, P.-C., Tseng, H.-E., Yang, S.-K., and Kuo, F.-H. (2021, January 8–10). New Multi-Access Network Transmission Technology to Enhance Edge Computing. Proceedings of the 2021 IEEE 22nd Asia-Pacific Network Operations and Management Symposium (APNOMS), Tainan, Taiwan.

4. (2024, July 01). A Quick Look at the Differences: RDMA vs. TCP/IP. Updated on 20 February 2023. Available online: https://community.fs.com/article/roce-vs-infiniband-vs-tcp-ip.html.

5. Mitchell, C., Geng, Y., and Li, J. (2013, January 26–28). Using One-Sided RDMA Reads to Build a Fast, CPU-Efficient Key-Value Store. Proceedings of the 2013 USENIX Annual Technical Conference (USENIX ATC 13), San Jose, CA, USA.

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3