New Cache Update Techniques to Reduce the Traffic on the Core Network

Nous redémarrons la nouvelle saison de séminaires le jeudi 1 octobre à 13h00. La présentation sera effectuée par Maggi Lorenzo, chercheur à "Huawei Technologies, Mathematical and Algorithmic Sciences Lab" et aura pour titre : "New Cache Update Techniques to Reduce the Traffic on the Core Network".

Content Distribution Networks (CDN) and Video on Demand applications use network caches to store the most popular contents near the user and reduce the traffic on the core network. The future projections for the cost of memory and bandwidth promote the use of caching to satisfy the ever increasing network traffic. Motivated by this, we study two new techniques to update the cache.The former part of the talk focuses on a new chunk-LRU cache replacement policy.
Rarely do users watch online contents entirely. Several online content platforms collect statistics on the user engagement performance of their videos, usually called “Audience retention rate”. We study how this information can be used to improve the performance of cache systems for video-on-demand and video-sharing platforms in terms of traffic reduction on the core network.
We first characterize the performance upper bound of a cache able to store parts of videos, when the popularity and the audience retention rate of each video are available to the cache manager. On the practical side, we then propose a chunk-LRU (Least Recently Used) cache management policy which operates on the first chunks of each video. We analyze its performance and provide a sufficient (and easily satisfied) condition under which refining the chunk granularity always improves the cache bandwidth saving performance.
The latter part of the talk formally studies a cache update technique exploiting the prediction of content popularity in the future.
The cache can only be updated at pre-defined off-peaks time instants. We formulate a Closed/Open-Loop (COL) problem that at each time decision instant takes into account the current content requests of users and the predicted future ones. In its original form, COL suffers from the curse of dimensionality. We  find a necessary and sufficient condition under which the COL problem simplifies drastically and only requires a limited prediction capability and a limited computational complexity.

Short Bio:
Lorenzo Maggi is a Researcher at Huawei Technologies, Mathematical and Algorithmic Sciences Lab. He graduated from University of Pavia (Italy) in 2008 and obtained his PhD Degree from EURECOM (Sophia Antipolis, France) in 2012. He has then been Researcher Assistant at University of Saarland (Germany) and Post-Doc fellow at Create-Net (Italy). He served as TPC-chair at Valuetools 2014 and he won the best paper award at WiOpt 2014. His main research interests are stochastic models and control for communication networks.

Lieu : salle 5

Horaire : jeudi 1 octobre à 13h00

Laboratoire Informatique d'Avignon

Université d'Avignon et des Pays de Vaucluse
339 chemin des Meinajaries, Agroparc BP 91228, 84911 Avignon cedex 9
+33 (0)4 90 84 35 00