Cocktail search in unstructured P2P networks

Resources search has become a very important area in light of the proliferation of peer-to-peer (P2P) networks and Grids, which heavily rely on effectiveness of forwarding algorithms to propagate queries among peers. Since none of any forwarding algorithms is perfect, Most existed search techniques exhibit varying performances when applying homogeneous forwarding logic in each peer. This paper proposes a promising approach to improve search through heterogeneous forwarding algorithms mixing, named Cocktail Search (CS) technique. Its core principle is that available forwarding algorithms are selected probabilistically in each peer to avoid the drawback of single algorithm. In this paper, the basic principle of CS technique is presented and Game theory is employed to verify its feasibility. The simulation results demonstrate that this technique shows significant advantages over others.