To improve unstructured P2P system performance, one wants to minimize the number of peers that have to be probed for the shortening of the search time. A solution to the problem is to employ a replication scheme, which provides high hit rate for target files. Replication can also provide load balancing and reduce access latency if the file is accessed by a large population of users. This paper briefly describes various replication schemes that have appeared in the literature and also focuses on a novel replication technique called Q-replication to increase availability of objects in unstructured P2P networks. The Q-replication technique replicates objects autonomously to suitable sites based on object popularity and site selection logic by extensively employing Q-learning concept.
View on arXiv