Analysis and Enhancements to Probabilistic Caching in Content-Centric Networking

In this paper, we first further investigate the use of random decision policies for caching schemes, or probabilistic caching, in Content-Centric Networking (CCN). Our main objective is to provide a mathematical model for the combination of random cache decision and least recently used (LRU) replacement policy in the content store of CCN, called LRU-PC (LRU with probabilistic caching). This analytical model contributes to the improvement on proper caching selection probability per content router, in order to achieve a good performance considering different parameters, such as the popularity of the files, cache size and others. Based on these results, the paper further contributes by extending the LRU-PC with a secondary list holding a wider view of the names of content packets that transverse a CCN node. In this case, a segment is stored in the cache with a given probability when its name is stored in this secondary list, dubbing this mechanism LRU-PCSL (LRU with probabilistic caching and a secondary list). As LRU-PC, LRU-PCSL has a constant time complexity, is scan-resistant, presents low memory requirements and is possible to implement in hardware. Evaluation results show that LRU-PCSL increases overall cache performance compared with LRU and LRU-PC.