Speaker: Jad Hachem
Affiliation: Ph.D. Candidate - UCLA
Abstract: The increase in mobile Internet usage has created a need to optimize content delivery in cellular networks. Traditional techniques to handle the high demand in wired networks consist in mirroring content across the network and placing it close to the end users. This is insufficient in wireless networks where the final hop is typically the bottleneck. To solve this problem, we consider the heterogeneous network architecture (HetNet) emerging in 5G that consists of a dense deployment of wireless access points (APs) with high data rate and small coverage combined with a sparse placement of cellular base stations (BSs) with large coverage and limited data rates. By placing caches at the APs and carefully designing what to store in them combined with what to transmit through the BSs, one can develop a framework to solve the content delivery problem in wireless networks. This talk explores the joint design of content caching and delivery, a recent technique known as “coded caching”, in these HetNets. The focus is on two main aspects: content popularity and network topology. The first part of this talk discusses the approximately optimal strategies to use when some files are much more popular than others, as is the case with for example YouTube videos. We show how different strategies are required depending on the density of users and the resulting load on the APs. We present the surprising result that in some cases it is inefficient to allocate as much memory as possible to the most popular files. In the second part of this talk, we discuss how the presence of multiple APs in the vicinity of each user can enhance the performance through various strategies. We also discuss the interference arising from the presence of multiple BSs, propose a separation-based strategy that separates the network and physical layers, and show that this separation architecture is approximately optimal.
Biography: Jad Hachem is a PhD candidate in the Department of Electrical and Computer Engineering at UCLA. His research interests include information theory and network information theory. For his doctoral research, he is working on content delivery in wireless networks. He has held internship positions at Qualcomm in 2015, where he worked on integrating vision and GNSS for enhanced positioning, and at the German Aerospace Center (DLR) in 2010 where he worked on dynamic co-operative positioning. He received the Roberto Padovani Scholarship for his work at Qualcomm. Jad received the BE degree in Computer and Communications Engineering from the American University of Beirut in 2011, and the MSc degree in Electrical Engineering from UCLA in 2013.
Date(s) - Nov 21, 2017
1:00 pm - 3:00 pm
E-IV Tesla Room #53-125
420 Westwood Plaza - 5th Flr., Los Angeles CA 90095