Improving Search in Unstructured P2P Systems: Intelligent Walks (I-Walks).

Random Walks (RW) search technique can greatly reduce bandwidth production but generally fails to adapt to different workloads and environments. A Random Walker can't learn anything from its previous successes or failures, displaying low success rates and high latency. In this paper, we propose Intelligent Walks (IW) search mechanism - a modification of RW, exploiting the learning ability and the shortest path distance of node neighbors. A node probes its neighbors before forwarding the query. The probe is to find a candidate that has the shortest distance from the query source and/or has ever seen before the object that is going to be sent. If there isn't such candidate, then a node is chosen as usual (at random). The experimental results demonstrate that new method achieves better performance than RW in terms of success rate. Keywords: Unstructured P2P, Search, Random Walks and Intelligent Walks. [ABSTRACT FROM AUTHOR]