Improved Flooding-Based Resource Discovery
Yadong Gong, Jiye Hu, Zhiliang Dong, Shengnian Wang, Shaojun Hu · 2010
One of key problems in peer-to-peer (P2P) systems is how to locate the interested resources. Most unstructured P2P networks utilize flooding-based resource discovery which assumes no knowledge about the network topology, and thus can give support to the node heterogeneity. These networks are highly resilient to nodes' failures, and incur low overhead at node arrivals and departures. However, flooding-based resource discovery makes such networks fail to obtain a good tradeoff between search results and search traffic. To address its inherent weaknesses, we propose a new flooding technique, called "Diff-Flooding", which takes node heterogeneity into consideration. In Diff-Flooding network, a query first try to hit the target by blindly visiting the nodes in local area to keep the search cost low, and if it fails to hit the target, it will resort to the nodes with high query answering capacities in large area to look for the target. We check the performance of our proposed search algorithm by comparing it with pure Flooding and Breadth-First-Search (BFS) mechanism. Simulation results indicate that, Diff-Flooding algorithm averagely saves 50 percent of search traffic while obtaining desirable search results.