Exploiting correlation among data items for cache replacement in Ad-hoc Networks

Ad-hoc Networks are special types of communication networks that don´t require any prior infrastructure to work. One of the crucial properties of such networks is their disconnected mode of operation. Due to the unpredictable nature of these networks, a node often has to work in offline mode with cached contents. Therefore, it is very vital in ad-hoc environment to determine the particular data to be stored in the cache and the position in the cache where that data has to be stored. In this paper, we have proposed a novel technique for cache replacement in Ad-hoc Network based on the mining of Association Rules. Using FP-Growth Association Rules Mining, the correlation among data items is computed and is then used as an important heuristics during cache replacement. We have evaluated the proposed technique using the simulator JIST/SWANS. A query routing algorithm for MANET is proposed and a simulation model is developed to analyze the proposed algorithm. The proposed cache replacement technique has been tested on different MANET configurations. The results reflect significant improvement in cache hit ratio with the proposed technique.