Computer Sciences and Information Technology
Title : | Caching Networks: an Information/Coding and Learning-theoretic View |
Area of research : | Computer Sciences and Information Technology |
Focus area : | Theoretical Sciences |
Principal Investigator : | Dr. Nikhil Karamchandani, Indian Institute Of Technology Bombay (IITB), Maharashtra |
Timeline Start Year : | 2023 |
Timeline End Year : | 2026 |
Contact info : | nikhilk@ee.iitb.ac.in |
Details
Executive Summary : | The last decade or so has seen an unprecedented growth in multimedia internet traffic, which in turn has put the underlying communication infrastructure under immense load. One proposed solution to mitigate this effect is to deploy caches at multiple locations in the network and use them to prefetch some of the popular content locally. Motivated by this, we propose to study the performance of content caching and delivery networks through an information/coding and learning-theoretic lens. Our focus will primarily be on the effect of the underlying network topology on the performance of the caching and delivery system. We plan to study the problem from two different viewpoints: 1) Information / Coding theoretic view: We consider a system consisting of users who are interested in accessing the content stored at a remote server. To improve the quality of service to the users, caches are distributed close to them and are populated with some of the content. Each user can access some subset of caches, based on its location for example, and the server can also assist in delivery via a shared link. The server transmission link is the bottleneck and the goal is to utilize the caches in such a way that the server transmission size is minimized. In the literature, special instances of this problem have been studied under the moniker “multi-access coded caching”. Achievability results have been proposed based on coding-theoretic ideas and some information-theoretic converses have also been derived. Our goal in this project is to advance the state of the art by closing the current gaps between the achievable rates and the converse results in some of the specific topologies studied in the literature, as well as developing new tools and techniques to devise coding schemes as well as converse results for more general connectivity topologies. 2) Learning-theoretic View: While the model above considered cache placement and delivery in phases, here we consider a situation where requests arrive sequentially and the cache contents can be updated in an online fashion, based on prior requests and caching decisions. Two popular metrics for evaluating the performance of online schemes are the competitive ratio and the regret. Within this framework, we plan to study the impact of network topology, request classes, and cache update costs on the design and analysis of optimal online caching and delivery schemes. |
Total Budget (INR): | 6,60,000 |
Organizations involved