Turning Memory into Bandwidth via Wireless Edge Caching: Fundamental Limits and Practical Challenges
Video is responsible for 66% of the 100x increase of wireless data traffic predicted in the next few years. Traditional methods for network capacity increase are very costly, and do not exploit the unique features of video. This talk gives a survey of a novel transmission paradigm based on the following two key properties: (i) video shows a high degree of asynchronous content reuse, and (ii) storage is the fastest-increasing quantity in modern hardware. Based on these properties, we suggest caching at wireless edge, namely, caching in helper stations (femto-caching) and/or directly into the user devices. We study two fundamentally different network structures: shared link caching networks and device-to-device (D2D) caching networks.
First, we present results based on network coded multicast delivery and/or D2D transmissions that show a “Moore’s law” for throughput: namely, in a certain regime of sufficiently high content reuse and/or sufficiently high aggregate storage capacity (sum of the storage capacity of all the users) in the network, the per-user throughput increases linearly, or even super-linearly with the cache size, and it is independent of the number of users for large network size, despite the fact that these users make independent and individual video files requests, i.e., the system does not exploit the naive broadcasting property of the wireless medium to send the same source to everybody. On the other hand, for both considered networks, we also provide information theoretic converse, by using which, we show that the proposed schemes achieves the order-optimal capacity. Then, we present the practical challenges and limitations of the achievable schemes. To overcome these challenges, for both network structures, we design novel polynomial-time complexity algorithms, which achieves near optimal performance such that they preserve the promised “Moore’s law” for throughput under realistic network parameter regimes.
Mingyue Ji is a final year PhD candidate at Ming Hsieh Department of Electrical Engineering, University of Southern California (USC). His adviser is Professor Giuseppe Caire, and he is also very fortunate to collaborate with Professor Andreas Molisch during his PhD study. Prior to USC, he worked as a research engineer and finished his Master thesis at the Access Technologies and Signal Processing Group in Ericsson, Stockholm, Sweden. He also obtained his Master of Science (MS) Degree in Electrical Engineering at Royal Institute of Technology (KTH), Sweden, and obtained his Bachelor Degree in Communication Engineering at Beijing University of Posts and Telecommunications (BUPT), China.