Smart Search over Desirable Topologies: Towards Scalable and Efficient P2P File Sharing

Abstract

Gnutella-like peer-to-peer networks exhibit strong small-world properties and power-law node degree distributions. However, the existing flooding-based query algorithms used in such overlay networks, knowing very little about these inherent natures, scale poorly with inefficient search and heavy traffic load, which is always a challenging problem to solve. In this paper, we eye our viewpoints upon the role of overlay topology in the search performance and propose a novel solution towards scalable and efficient peer-to-peer distributed file sharing, by making better use of such emergent topological properties of these networks. We first, by examining what inspirations can be taken from these properties, provide several helpful guidelines as the design rationale of our solution, and we then propose a new technique for constructing DesirableTopologies and a novel Smart Search algorithm operating on them, as two key components of our solution. To justify the performance gains of our techniques, we also conduct extensive experiments under realistic network conditions and make an all-around comparison with currently well-known systems.