Abstract—Many researchers have observed that TCP’s congestion control mechanisms can lead to bursty traffic flows on modern high-speed networks, with a negative impact on overall network efficiency. A proposed solution to this problem is to evenly space, or “pace”, data sent into the network over an entire round-trip time, so that data is not sent in a burst. In this paper, we quantitatively evaluate this approach. Pacing offers better fairness, throughput, and lower drop rates in some situations. However, we show that contrary to intuition, pacing often has significantly worse throughput than regular TCP because it is susceptible to synchronized losses and it delays congestion signals. We propose and evaluate approaches for eliminating this effect.
Amit Aggarwal, Stefan Savage, Thomas E. Anderson