Round-Robin Streaming with Generations

TL;DR

This paper analyzes different application layer coding schemes for streaming over lossy links, comparing their efficiency, energy consumption, and throughput through analytical and experimental methods, highlighting the trade-offs based on generation size and coding complexity.

Contribution

It provides a comprehensive comparison of coding schemes for streaming, including analytical models and experimental validation, and identifies optimal coding strategies based on generation size and device constraints.

Findings

Coding improves net throughput and reduces energy consumption.

Heavy coding causes packet drops on resource-constrained devices.

Systematic random linear coding is optimal for larger generations.

Abstract

We consider three types of application layer coding for streaming over lossy links: random linear coding, systematic random linear coding, and structured coding. The file being streamed is divided into sub-blocks (generations). Code symbols are formed by combining data belonging to the same generation, and transmitted in a round-robin fashion. We compare the schemes based on delivery packet count, net throughput, and energy consumption for a range of generation sizes. We determine these performance measures both analytically and in an experimental configuration. We find our analytical predictions to match the experimental results. We show that coding at the application layer brings about a significant increase in net data throughput, and thereby reduction in energy consumption due to reduced communication time. On the other hand, on devices with constrained computing resources, heavy coding operations cause packet drops in higher layers and negatively affect the net throughput. We find from our experimental results that low-rate MDS codes are best for small generation sizes, whereas systematic random linear coding has the best net throughput and lowest energy consumption for larger generation sizes due to its low decoding complexity.