Scholarlink

As data center network topologies become increasingly complex and network traffic continues to surge, congestion control has emerged as a formidable challenge. Concurrently, the utilization of lossless networks can lead to issues such as PFC storms and PFC deadlocks. Our research indicates that existing end-to-end congestion control protocols are insufficient in fully utilizing bandwidth or responding promptly to congestion. In this paper, we introduce RVCC, a PFC-based hop-by-hop congestion control algorithm designed to address the adverse effects of PFC in lossless networks, thereby optimizing bandwidth utilization. RVCC achieves rapid and fine-grained congestion control through queue selection and switch communication. Experimental results obtained from the NS3 simulation demonstrate that RVCC significantly reduces tail latency and average latency compared to state-of-the-art algorithms.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

RVCC: Congestion Control to Reduce Victim Flows in Data Center Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Lang Dai,
  • Lu Tang

摘要

As data center network topologies become increasingly complex and network traffic continues to surge, congestion control has emerged as a formidable challenge. Concurrently, the utilization of lossless networks can lead to issues such as PFC storms and PFC deadlocks. Our research indicates that existing end-to-end congestion control protocols are insufficient in fully utilizing bandwidth or responding promptly to congestion. In this paper, we introduce RVCC, a PFC-based hop-by-hop congestion control algorithm designed to address the adverse effects of PFC in lossless networks, thereby optimizing bandwidth utilization. RVCC achieves rapid and fine-grained congestion control through queue selection and switch communication. Experimental results obtained from the NS3 simulation demonstrate that RVCC significantly reduces tail latency and average latency compared to state-of-the-art algorithms.