A Cooperative Cache Replacement Algorithm Using Machine Learning For Information Centric Mamet

Y.J.Sudha Rani, M. Seetha

The restrictions of the memory size on the MANET device nodes have intrigued the need for efficient cache management. The conspicuous intrigue of MANETs is that the system is decentralized, and hubs or gadgets are versatile, in other words there is no fixed foundation which gives the likelihood to various applications in various territories, for example, natural checking, calamity alleviation and military correspondences. Since the mid-2000s enthusiasm for MANETs has incredibly expanded which, to some extent, is because of the reality portability can improve arrange limit. The ill management of the cache memory can lead to the slowdown of the applications running on the MANET. Also, frequent replacements of the cache data can increase the time complexity or decrease the responsiveness of the applications running on the network nodes. To overcome this bottleneck a number of research attempts were carried out and these attempts were criticised for higher time complexity due to the replacement strategies used in those research outcomes. It is to be observed that the cooperative caching is gaining the popularity for building efficient MANET architectures due to various benefits. Nevertheless, none of the parallel research outcomes have considered the benefits from this architectural strategy. This work builds a novel technique to reduce the cache miss situations using the cooperative cache sets in the network. The proposed algorithm relies on the routing table for building the association between the nodes under a cooperative cache structure. From the cooperative structure, the proposed method identifies the node in the network with the searching data items and based on the adaptive threshold, the data is supplied to the demanded node. Further, using an extension of the proposed algorithm, this work replaces the cache data in the victim cache location using a machine learning based prediction. Finally, the result from the proposed algorithm demonstrates nearly 95%cache hit rate, with nearly 15% and 99% improvement in time complexity over parallel research outcomes.

Volume 12 | Issue 2

Pages: 1379-1388

DOI: 10.5373/JARDCS/V12I2/S20201177