Pcoord: Network position estimation using peer-to-peer measurements (original) (raw)

Several recently emerged Internet services make use of application-level or overlay networks. Examples of such services include overlay multicast, structured peer-to-peer lookup services, and peer-to-peer file sharing. Many of these services could benefit from a network distance prediction mechanism that estimates inter-host latencies. In this paper, we present PCoord, a distributed network coordinate system for overlay topology discovery and distance prediction. PCoord assigns coordinates to hosts on the Internet so that the Euclidean distances between hosts' coordinates accurately predict their network latencies. Most of the existing network coordinate systems rely on a fixed set of landmark nodes for coordinate computation. PCoord, in contrast, is a fully decentralized system that allows participating hosts in an overlay network to collaboratively construct an accurate geometric model of the overlay topology using a small number of peer-to-peer measurements. Under the PCoord framework, we present three different peer sampling and coordinate mapping schemes: RandPCoord, ClusterPCoord, and ActivePCoord. Through extensive simulations using both real network measurements and simulated topologies, we show that the geometric model constructed by PCoord predicts pair-wise host distances accurately and efficiently.