Adaptive suppression of inter-packet delay variations in coded packet networks

Network coding tends to increase not only the end-to-end delay due to unavoidable buffering of packets, but it also increases the inter-packet delay variations (jitter) due to batching. While an increase in delay does not necessarily adversely affect throughput, sudden delay spikes may interfere with TCP´s congestion avoidance mechanism. Such spikes are a common phenomenon in coded packet networks since packet loss may prevent the receiver from decoding a whole generation until sufficient redundant packets have arrived. After successful decoding, a burst of packets is forwarded. In this paper, we propose an adaptive queueing-based approach to minimize the inter-packet delay variation by dynamically delaying decoded packets at the receiver side depending on the current suppression queue backlog, average delay, and packet loss rate. We evaluate our approach by measuring TCP throughput on top of an RLNC system over a lossy wireless link. The TCP throughput significantly improves when our proposed method is used for interpacket delay variation suppression. The method is implemented as a lightweight library and will be made available for download under the GNU GPLv2 at [1].