Optimizing Multi-path QUIC Protocol Performance in Cloud-Edge Collaboration through CPU Affinity and Kernel-Level Resource Scheduling

被引：0

作者：

Xia, Xiaohan ^{[1
]}

Chen, Bin ^{[1
]}

Cai, Chao ^{[1
]}

Gao, Pei ^{[1
]}

Qiu, Jiahui ^{[1
]}

Hou, Yinglong ^{[1
]}

机构：

[1] China United Network Commun Co Ltd, Intelligent Network Innovat Ctr, Beijing, Peoples R China

来源：

2024 IEEE/CIC INTERNATIONAL CONFERENCE ON COMMUNICATIONS IN CHINA, ICCC | 2024年

关键词：

Multi-path QUIC; CPU affinity; Kernel-level Resource Scheduling; Cloud-Edge Collaboration; eBPF;

D O I：

10.1109/ICCC62479.2024.10681738

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

This study investigates how optimizing CPU affinity and kernel-level resource scheduling can enhance the performance of the Multi-path QUIC (MPQUIC) protocol in cloud-edge collaboration. It explores the significance of CPU affinity for improving MPQUIC protocol performance and leverages extended Berkeley Packet Filter (eBPF) technology to implement a heuristic planning algorithm for achieving high-performance implementations. By adjusting CPU affinity, it effectively enhances the maximum throughput achievable by MPQUIC on a single core and the relationship between acceleration bandwidth and CPU consumption. Furthermore, combining eBPF technology's Scheduling algorithm further optimizes the protocol performance in cloud-edge environments. This research provides a new perspective and approach for optimizing the performance of Cloud-Edge Collaboration using MPQUIC.

引用

页数：6