|
|
|
|
|
|
|
High-bandwidth content dissemination is a first class application running over the Internet, ranging from bulk data transfer to video-on-demand and software updates dissemination. For scalability reasons, the peer-to-peer paradigm has been widely used raising various interesting challenges. The talk will focus on two specific aspects of collaborative high-bandwidth content dissemination systems: dealing with selfish users and coding techniques. The first part of the talk focuses on the problem of free-riding, i.e., peers that do not contribute their fair share to the system, and proposes LiFTinG, a lightweight (possibly colluding) free-riders tracking protocol in gossip designed for epidemic content dissemination systems such as most streaming protocols deployed in the public domain. LiFTinG is composed of direct and a posteriori verifications involving logs and cross-checking, and leverages the dynamic and the randomness of epidemic protocols to prevent collusion. The second part of the talk focuses on network coding techniques and presents LTNC, low-complexity networks codes based on the Luby transform. In its current form, namely random linear, network coding has high decoding costs which prevent its use on limited capabilities devices such as cell-phones and sensors. At the core of LTNC lies a set of heuristics to efficiently build fresh encoded blocks from LT-encoded ones (which involves NP-complete problems) while preserving the LT structure of the code. Bio: Kévin Huguenin received his B.Sc. of computer science from the Université of Rennes, France, in 2005 and his M.Sc. from Ecole Normale Supérieure de Cachan and the Université de Nice - Sophia Antipolis, France, in 2007. During his studies, he visited the Vrije Universiteit Amsterdam for 8 months. He obtained his Ph.D. from Université de Rennes, in the ASAP research group, under the guidance of Anne-Marie Kemrarrec. Starting March 2011, he will work as a post-doctoral fellow with Bettina Kemme at McGill, Canada. His research interests include misbehavior detection in large-scale distributed systems with a special attention to content dissemination systems, on-line social networks and massively multi-player games.
|
|
|
|
|
|
|
|