Afuronto: A Six Hop Peer-to-Peer Network of Zillions of Nodes

Abstract

We present, in this paper, a new overlay network design that is suited for large scale peer-to-peer networks requiring fast message transmission. We particularly address the problem of designing an overlay communication network that spans across a set of data centres located in multiple geographic locations with the necessity of requiring locality properties for applications. Our design for the overlay network is motivated by the social network structures and we use different operators as found in evolutionary computing to emulate a semi-structured network. We present the network structure and the various algorithms required for basic communication and maintenance. We also study the proposed network by simulation and compare with existing overlay networks. The simulated results show that on an average there are at most six hops required for a message originated at any node in the network to be transferred to any other node in the network. The network so designed is scalable, and resilient to failure and, can include node of arbitrary in processing capacity, network bandwidth and storage.