Information-centric networks (ICN) is a future Internet architecture that rearchitects the current host-centric Internet to a content-centric one. Caching content within the intermediate nodes is one of the salient features of ICN. This in-network caching allows the content requests to be served from the intermediate nodes rather than the origin servers, thus reducing the content access time and the load on servers. Existing literature proposes many caching strategies for ICN and Leave Copy Everywhere (LCE), Leave Copy Down (LCD), Cache Less for More (CL4M) and ProbCache are the most popular ones. Performance of caching strategies vary significantly according to the behavior of the underlying network nodes. We evaluate the performance of the aforementioned caching strategies in diverse network settings and analyze which strategy is most suitable in specific scenarios. In this work, we consider static networks, synthetic mobile networks, and real-world pedestrian and vehicular mobile networks. Specifically, we consider static academic networks (WIDE, GEANT, GARR), two synthetic mobility models – grid and random waypoint, a pedestrian network designed using Stockholm pedestrian trace, and vehicular networks designed using Rome taxicab trace and Seattle bus trace. We conduct experiments in Icarus, a simulator extensively used for ICN research, using YouTube access trace, a real-world request stream trace.
