On the implementation of robust distributed peerspaces

Abstract

With respect to temporal and spatial coupling properties, coordination models based on the persistent tuple spaces can be easily integrated with peer-to-peer infrastructure for scalability and robustness. With a system that forms a persistent shared space model on top of richer P2P overlay, issues that crop up with tightly coupled distributed shared space models can be eliminated. Tuple space models on top of free-Pastry [1] overlay addresses previously mentioned issues well because of overlay networks have in-built functionalities plus inherited features such as decentralization, structured and self organization. In our work, a tuple space framework has been developed on top of pastry which is a DHT [3] based implementation and is highly scalable, and also uses its predefined functions to join the network and routing for locating tuples. Moreover, in our approach the middleware provides an interface to access tuple space in order to execute jobs using primitives such as take (), read() and write(). This tuple space model is a NoSQL abased approach since it uses column based mechanism for searching [4].