Enabling efficient information discovery in a self-structured grid

One of the key success factors enabling the deployment of large scale grid systems is the existence of efficient resource discovery mechanisms. Accordingly, the main issues to be addressed by such a grid information system are those of scalability and minimal network overhead. In this respect, we propose a solution based on proactive information caching supported by a self-structured overlay topology. The proposed approach features a fully distributed ant-inspired self-organized overlay construction that maintains a bounded diameter overlay, and a selective flooding-based discovery algorithm that exploits local caches to reduce the number of visited nodes. To improve the caching scheme while retaining minimal bandwidth consumption, cache contents are periodically exchanged between neighboring nodes using an epidemic replication mechanism that is based on a gossiping algorithm, thus allowing nodes to have a more general view of the network and its resources. Extensive experimentation provides evidence that the average number of hops required to efficiently locate resources is limited and that our framework performs well with respect to hit rate and network overhead.