Broadcast Approach Meets Network Coding for Data Streaming

TL;DR

This paper introduces a layered adaptive causal random linear network coding scheme for data streaming that effectively balances throughput and delay, significantly improving performance over traditional non-layered methods.

Contribution

The work proposes a novel layered coding scheme (LAC-RLNC) that manages high-dimensional throughput-delay trade-offs with adaptive layering and error correction for streaming.

Findings

LAC-RLNC achieves up to three times lower delay for the base layer.

The layered approach significantly outperforms non-layered schemes in delay metrics.

Numerical results confirm near-optimal performance close to theoretical bounds.

Abstract

For data streaming applications, existing solutions are not yet able to close the gap between high data rates and low delay. This work considers the problem of data streaming under mixed delay constraints over a single communication channel with delayed feedback. We propose a novel layered adaptive causal random linear network coding (LAC-RLNC) approach with forward error correction. LAC-RLNC is a variable-to-variable coding scheme, i.e., variable recovered information data at the receiver over variable short block length and rate is proposed. Specifically, for data streaming with base and enhancement layers of content, we characterize a high dimensional throughput-delay trade-off managed by the adaptive causal layering coding scheme. The base layer is designed to satisfy the strict delay constraints, as it contains the data needed to allow the streaming service. Then, the sender can manage the throughput-delay trade-off of the second layer by adjusting the retransmission rate a priori and posterior as the enhancement layer, that contains the remaining data to augment the streaming service's quality, is with the relax delay constraints. We numerically show that the layered network coding approach can dramatically increase performance. We demonstrate that LAC-RLNC compared with the non-layered approach gains a factor of three in mean and maximum delay for the base layer, close to the lower bound, and factor two for the enhancement layer.