SQCC: Stable Queue Congestion Control
Jian Tang,Tingting Xu,Liming Wang,Zhenjie Lin,Ming Zhao,Xiaoliang Wang,Cam-Tu Nguyen,Chen Tian,Zhuzhong Qian,Wenzhong Li
DOI: https://doi.org/10.1109/nof58724.2023.10302751
2023-01-01
Abstract:The advent of high-speed networks has revolutionized data center capabilities by providing low latency and high bandwidth for applications. However, traditional TCP congestion control algorithms are no longer adequate for data center networks. RTT-based congestion control leverages advanced NIC hardware to identify accumulated queuing delay of the end-to-end path. It is simple, effective, and adaptable to different environments. Nevertheless, RTT-based congestion control faces challenges related to unstable queue length and oscillation caused by RTT feedback delays. With the increase of queue length, the oscillation range also amplifies. To address these issues, we propose SQCC, which introduces two key enhancements. Firstly, it employs a novel error function to regulate the queue length within a controlled range that is proportional to the number of incast flows. Secondly, it incorporates self-adjustable parameters for rate increment and RTT threshold, effectively managing queue oscillation and ensuring a non-empty link. We evaluate the algorithm’s effectiveness through NS3 network simulations, and the results demonstrate that SQCC achieves an 80% reduction in queue size upon convergence and exhibits a significantly low oscillation range (27% to 57%) in large-scale incast scenarios.