A Comparative Study of In-network Caching Strategies for Information Centric Networks

Abstract

One of the most efficient methods to decrease the overall network traffic is through the deployment of distributed caching systems. Thus, are many proposed strategies to determine where to locate the caches within a network to maximize its performance. However, among all these strategies, it is still unclear, which one performs better. In this paper, the in-network caching strategies for Information Centric Networks (ICN) are studied. A neutral simulation framework is used to measure the performance in terms of cache hit ratio, content delivery latency, and the generated link load. The study considers different cache positioning strategies i.e., Leave Copy Everywhere (LCE), Leave Copy Down (LCD) and Probabilistic Caching (Probcache), different content popularities range, a variety of cache sizes and a couple of eviction policies i.e., Least Recently Used (LRU) and Least Frequently Used (LFU) algorithms. Results show that when LRU implemented as an eviction policy, the ProbCache performs better than LCD by average of 3% on the cache hit ratio when the cache size is small i.e., (.002,.004) with height content popularity value i.e., (.8, 1, 1.2) and better than LCE by average of 27% where LCD performs better in bigger cache size i.e., (.01,.05). Furthermore, we concluded with that, if we implement LFU as an eviction policy, the LCE and LCD strategies are the best strategy that fit’s the environment regarding the cache hit ratio and latency. Additionally, it is noticeable when LFU is implemented the cache hit ratio increases by average of 20% from LRU.