Scholarlink

<p>In early 2017, Google announced a new version of the TCP congestion control algorithm called Bottleneck Bandwidth and Round-trip Time (BBR), aimed at eliminating bottleneck queue formation and optimizing the utilization of available bottleneck bandwidth on networks. Despite the remarkable throughput and delay improvements demonstrated by BBR on modern networks, it exhibits deficiencies and experiences suboptimal delay performance, primarily on network paths featuring TCP acknowledgment (ACK) compression and aggregation. This is due to one of its underlying design choices. ACK compression and aggregation result in excessive ACK rates. However, BBR relies on a traditional approach that cannot thoroughly filter these excessive recorded ACK rates. This leads to BBR injecting additional bytes of data into the buffer queue, causing delivery rate overestimation, and, consequently, prolonged queueing delays. This study proposes the BBR with Enhanced Bandwidth Estimation (BBR-EBE+) algorithm to improve BBR’s bandwidth estimation by accurately detecting and filtering excessive ACK rates and delaying the update of the estimated bandwidth. We implemented our experimental evaluations in a physical testbed environment. BBR-EBE+ achieved up to a 28% average reduction in queueing delay, increasing to 36% in worst-case scenarios, along with an average throughput increase of about 12% compared to other recently published algorithms, such as calibrating bandwidth estimation (CBE).</p>

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

BBR-with enhanced bandwidth estimation (BBR-EBE+): an improved BBR congestion control algorithm based on TCP acknowledgment compression and aggregation

  • 期刊论文

  • 发表日期：

    2025年2月21日

查看全文
  • Charles Kihungi Njogu,
  • Wang Yang,
  • Humphrey Waita Njogu,
  • Adrian Bosire

摘要

In early 2017, Google announced a new version of the TCP congestion control algorithm called Bottleneck Bandwidth and Round-trip Time (BBR), aimed at eliminating bottleneck queue formation and optimizing the utilization of available bottleneck bandwidth on networks. Despite the remarkable throughput and delay improvements demonstrated by BBR on modern networks, it exhibits deficiencies and experiences suboptimal delay performance, primarily on network paths featuring TCP acknowledgment (ACK) compression and aggregation. This is due to one of its underlying design choices. ACK compression and aggregation result in excessive ACK rates. However, BBR relies on a traditional approach that cannot thoroughly filter these excessive recorded ACK rates. This leads to BBR injecting additional bytes of data into the buffer queue, causing delivery rate overestimation, and, consequently, prolonged queueing delays. This study proposes the BBR with Enhanced Bandwidth Estimation (BBR-EBE+) algorithm to improve BBR’s bandwidth estimation by accurately detecting and filtering excessive ACK rates and delaying the update of the estimated bandwidth. We implemented our experimental evaluations in a physical testbed environment. BBR-EBE+ achieved up to a 28% average reduction in queueing delay, increasing to 36% in worst-case scenarios, along with an average throughput increase of about 12% compared to other recently published algorithms, such as calibrating bandwidth estimation (CBE).