Evaluating Unstructured Peer-to-Peer Lookup Overlays ABSTRACT

Unstructured peer-to-peer lookup systems incur small constant overhead per single join or leave operation, and can easily support keyword searches. Hence, they are suitable for dynamic failure-prone environments. In this paper, we define metrics for evaluating unstructured overlays for peer-to-peer lookup systems. These metrics capture the search dependability and efficiency, and the granularity at which one can control the tradeoff between the two, as well as fairness. According to these metrics, we evaluate different graphs and overlays, including a Gnutella graph, a power law random graph, normal random graphs, a 3-regular random graph, and a 3-Araneola overlay. Our study shows that, according to our metrics, a 3-Araneola overlay achieves the best results, and hence it is an excellent solution for flooding-based peer-to-peer lookup system