Geographically-Distinct Request Patterns for Caching in Information-Centric Networks

The current Internet architecture follows a hostcentric communication model, intended for machine to machine connection and message passing. Modern Internet users are mainly interested in accessing information by name, irrespective of physical location. Information centric networking (ICN) was developed to rethink Internet foundations. Innetwork caching is one of the main features of ICN. Studying the performance of different caching algorithms in ICN requires a good understanding of users' request distributions in such networks. Most studies use simplifying assumptions for user request patterns since ICNs are not yet deployed. Geographically localized and global request patterns have both been observed to possess Zipf-like properties, although the local distributions are poorly correlated with the global distribution. Several independent Zipf distributions combine to form an emergent Zipf distribution in real client request scenarios. We develop an algorithm that can generate realistic synthetic traffic for geographic regions that possesses Zipf power-law properties as well as a global Zipf distribution. Our simulation results show that the caching performance would have different behaviour based on users' requests distribution.