With the recent explosive growth in online classes and virtual meetings, real-time video communication has quickly become essential to everyday life. Despite its widespread deployment, our investigation revealed that current protocols, ranging from industry standards such as WebRTC to state-of-the-art research such as Salsify, frequently perform sub-optimally in the presence of competing flows at the same bottleneck. For example, WebRTC’s throughput can degrade from 73% to a mere 8% of the available bandwidth when competing with just two TCP flows. We tackle this problem in this work by introducing a novel PIBES protocol for real-time video applications to operate in the presence of competing TCP traffic. PIBES employs a new inband bandwidth estimation method that can quickly and accurately measure the bottleneck link bandwidth even with competing flows. Moreover, PIBES can detect the absence or presence of competing flows, which enables it to maximize video quality when there is no competing flow and to maintain acceptable video quality while sharing bandwidth with competing flows. Experiments demonstrate that PIBES achieves throughput and delay comparable to the state-of-art protocols, but outperforms them significantly in the presence of competing TCP flows.

中文翻译：

---

PIBES-实时视频应用的竞争流感知协议

随着在线课程和虚拟会议的爆炸性增长，实时视频通信已迅速成为日常生活中必不可少的部分。尽管已广泛部署，但我们的调查显示，从行业标准（例如WebRTC）到最新研究（例如Salsify），当前的协议在存在相同瓶颈的竞争流量的情况下常常表现欠佳。例如，仅与两个TCP流竞争时，WebRTC的吞吐量就可以从可用带宽的73％降低到仅为8％。在这项工作中，我们通过为实时视频应用引入新颖的PIBES协议以在竞争TCP流量的情况下运行来解决此问题。PIBES采用了一种新的带内带宽估计方法，即使在竞争流的情况下，该方法也可以快速而准确地测量瓶颈链路带宽。此外，PIBES可以检测到竞争流的不存在或存在，这使得它可以在没有竞争流的情况下最大化视频质量，并在与竞争流共享带宽的同时保持可接受的视频质量。实验表明，PIBES可以实现与最新协议相当的吞吐量和延迟，但是在存在竞争的TCP流的情况下，PIBES的性能和性能明显优于它们。