In this thesis, we proposed a Cluster-Based Cache Replacement (CBR) scheme for 5G Networks to reduce the backhaul traffic. We developed our scheme based on the understanding of the degradation of the performance of the cache placement algorithm. We expect that whenever file request pattern differs from the file popularity distribution, such as unpopular files become more popular or vice versa, the caching system should experience performance degradation. We address this problem by presenting a cache replacement scheme based on the idea of Least Frequency Used (LFU) replacement policy, but we consider only the recent request to avoid cache pollution. We evaluated the performance of CBR through simulation and compared its performance with LRU that is widely used as a cache replacement technique in practice. We simulated three different configurations of LRU scheme in a cluster-based mobile network model. Our simulation results show that the CBR outperforms LRU, where it reduces the miss ratio from 86% to 76% and the backhaul traffic from 3.67×105 to 3.47×105 MB with 10% of cache size. This superior performance it achieves by fewer replacement decisions and storing more files in the cache.
What is a good buffer cache replacement scheme for mobile flash storage?
