Selective Walk Searching Algorithm for Gnutella Network
Yan Lei Xu, Xiaojun Ma, Charles Wang · 2007
Locating a resource or service efficiently in unstructured decentralized peer-to-peer systems is one of the most important issues for peer-to-peer applications. We propose a selective walk searching method for Gnutella systems, which is based on the random walk algorithm in conjunction with emphasis on the peers' selection criteria. The selection is based on several hints of neighboring peers, including the number of the shared files, the most recent succeeded queries and the specific peer that provided the answer. By using these hints, a peer selection mechanism is proposed. The searching source selects a set of neighbors according to the selection criteria. In addition, a flow control mechanism is adopted in order to avoid the overload of hotspots and message loss. The flow control is based on the rate limit and message priority. Simulation results show that the selective walk searching algorithm outperforms the pure random walk algorithm in most evaluation metrics especially query success rate.