Abstract

This talk presents the background to our interest in optimising modern streaming behaviour over new queue management strategies likely to emerge in last-mile networks in coming years, and some of the technical choices we have made to experimentally explore this space.

With Netflix and YouTube accounting for more than 50% of North American fixed network peak download traffic in 2016, multimedia streaming represents a significant source of Internet traffic. Dynamic Adaptive Streaming over HTTP (DASH) is a recent standard for live and on-demand video streaming services, where clients adapt their behaviour on-the-fly to match regularly updated estimates of network capacity. Consumer DASH streams are likely to be bottlenecked by emerging Active Queue Management (AQM) schemes such as Proportional Integral controller Enhanced (PIE) and FlowQueue-Controlled Delay (FQ-CoDel) which are being progressively deployed either at the home gateway or at the Internet Service Providers’ end points in order to counter bufferbloat.

We study the intricate interactions between DASH streams and AQMs and propose a new use case – using intra-chunk parallel network connections (chunklets) for DASH content retrieval over AQM bottlenecks. We experimentally evaluate and characterise the impact of using chunklets over traditional FIFO , PIE and FQ-CoDel bottlenecks. We show that chunklets enable the DASH client to attain a throughput multiplication effect, hence translating to better user experience in the presence of competing elastic flows.

Biography:

Jonathan Kua received the Bachelor of Engineering (First Class Hons.) degree in telecommunications and network engineering from Swinburne University of Technology, Melbourne, Australia in 2014. He is currently a PhD candidate at the Internet For Things (I4T) Research Laboratory within the School of Software and Electrical Engineering at Swinburne University of Technology. His research interests include IP-based content delivery, adaptive multimedia streaming, data transport protocols and active queue management.