Frozen Cache: Mitigating Filter Effect and Redundancy for Network of Caches

Abstract

Information-Centric Networking (ICN) architecture leverages the network of caches’ idea to bring content closer to consumers to ultimately reduce the load on content servers and prevent unnecessary packet re-transmissions. Nevertheless, the performance of existing cache management schemes mainly developed for a single cache is inadequate for a network of caches. There are many factors such as data dependencies, data redundancy, the limited size of caches, poor replacement policies, and many other factors that negatively impact a network of caches. Besides, traffic correlation among different caches on the network influences the performance of the network of caches. One of the essential correlations is the edge filtering effect. In the presence of data redundancy, the edge filtering effect even becomes more severe. The cache filtering effect happens when all arriving requests inspect the first cache for data. Therefore, the subsequent caches in the network receive only those requests that could not find data (cache-miss) from the edge cache. In this paper, we propose Frozen-cache to mitigate the filtering effect. This policy repeatedly freezes content in a cache to allow subsequent caches to receive popular content. A lightweight coordinated scheme incorporated with Frozen-cache policy to cope with the data redundancy problem. Based on our experiments obtained from realistic scenarios, the Frozen-cache idea highly outperforms state of the art caching schemes. Depending on the network setup, this superiority varies from 25% to 700%.