Per-CCA Queueing

Yara Mulla, Isaac Keslassy

Abstract: Due to their increasing aggressiveness, recent congestion control algorithms (CCAs) can starve vanilla TCP flows in their shared router queues. Unfortunately, existing router-based solutions cannot prevent this starvation.In this paper, we introduce a per-CCA queue isolation where incoming flows first undergo CCA classification, and then are mapped to isolated queues based on their classified CCA. We provide a fundamental analysis for this per-CCA isolation, and present two models for its performance. Then, in evaluations, we show how this per-CCA isolation clearly outperforms buffer sharing, and how our advanced model can accurately represent its performance. We further show how we can increase fairness by optimizing the queue service rates.