Dream

Abstract

In this paper we present a distributed framework that supports customized deployment of a variety of indexing engines over million-node overlays. The key aim is to provide the appropriate integrated set of tools that allows numerous applications with large-scale, different requirements to evaluate and test the performance of various application protocols for very large scale deployments (multi million nodes -billions of keys). Using lightweight and efficient collection mechanisms, our system enables real-time registration of multiple measures, integrating support for real-life parameters such as node failure models and recovery strategies. Experiments have been performed at the PlanetLab network and at a typical research laboratory in order to verify scalability and show maximum re-usability of our setup.