Hermann Schloss

Abstract—Topological distance estimation is the key to the efficiency in distributed systems and peer-to-peer networks. Contrary to many existing or proposed methods, which usually require the exchange of messages between the nodes, we have developed a metric, which is computed purely within a node, and which is based on the preloaded and precomputed topological structure of the Internet. Many distributed systems and applications may benefit from this metric, since it estimates the topological distance between any arbitrary pair of nodes in the Internet. As a proof of concept we have first shown the correlation between our metric and a few established distance indicators, such as hop count or round trip time of a message. Then, we employed this metric as an “edge weight” representing the connection quality between two network nodes and we used it for the construction of a multicast overlay network based on a Minimum Spanning Tree approximation. According to the evaluation results, this metric corresponds fairly well to the actual measured distances. By using this met-ric, our approach minimizes communication costs and avoids extraneous communication needed for latency measurements. Keywords-Internet, Topology, Distance Estimation Method I