A novel self-adaptive search algorithm for unstructured peer-to-peer networks utilizing learning automata

Designing an efficient search algorithm is an important issue in unstructured peer-to-peer networks when there is no central control or information on the locations of objects. There are various search strategies with different effects on network performance. In k-random walks as a search strategy, having an adaptive value of k instead of a random value can affect performance of the network. Therefore in this paper, a distributed novel self-adaptive search algorithm has been developed by application of learning automata to overcome this challenge. This method does not aim to determine the value of k for k-random walks algorithm and each peer can issue walkers in a self adaptive manner. Simulation results show that the proposed search algorithm improves some features such as average number of walkers per query, average number of produced messages, number of hits per query and also success rate efficiently in comparison with the k-random walks algorithm.