Supervision : Pierre SENS

Co-supervision : SIMON Gwendal, MONNET Sébastien

Delivery, Transcoding and Neutrality on Large Live Streaming Systems

Today many devices are capable to capture full HD videos and to use its network connection to access Internet. The popularization of devices and continuous efforts to increase network quality has brought a proper environment for the rise of live streaming. Associated to the large scale of Users Generated Content (UGC), live streaming presents new challenges. In this thesis we target the delivery and transcoding of live streaming systems.
First, to study the aspects we target of live streaming systems we need to characterize them and evaluate the proposed solutions with relevant traces. Therefore our first contribution is a data set, and its analysis, containing three months traces of two UGC live streaming services.
Second, we explored and developed solutions for the delivery of the content produced by these platforms. One of the challenges is the heterogeneity among streams popularity, which generally implies over- provisioning and consequently resource waste. We show that there is a trade-off between the number of servers involved to broadcast the streams and the bandwidth usage among the servers. We also stress the importance to predict streams popularity in order to efficiently place them on the servers.
Lastly, we target the difficulties concerning transcoding of live streams. The transcoding operations over streams are computing consuming and are key operations on adaptive bit rate streaming. We show that adaptive streaming is able to reduce the delivery bandwidth cost and to increase viewer quality of experience at the cost of computing resources. We formulate two management problems to address the trade-off between benefits and costs.

2014-2015 Publications