Intelligent caching in information-centric networks
Ho Tin Chiu · 2021
Caching has become an integral part of the Internet to improve the performance of services and applications that rely on the Internet as a medium to disseminate information and content. Information-centric networks (ICN) is an alternative paradigm to reconstruct the Internet that puts information and content as the first class citizens. Under such paradigm, caching is natively supported in the network layer. In this thesis, we address the inefficiencies resulting from the default caching strategy employed by ICN forwarders, and discuss the design and implementation of two caching schemes that improve caching in the spatial and temporal dimensions. In the spatial dimension, we introduce cooperative caching. Forwarders share their caches with one-hop neighbors while evicting redundant cache entries. We make several design decisions to make this scheme practical to implement. In the temporal dimension, we explore the use of machine learning methods to construct a model that is used to predict future content popularity based on past history, and design a cache algorithm that uses predicted popularity to make cache decisions. We evaluate the caching schemes by extensive emulations of network traffic on emulated network topologies.