Resource Discovery and Selection for Large Scale Query Optimization in a Grid Environment

Current peer to peer (P2P) methods employing distributed hash tables (DHT) for resource discovery in a Grid environment suffer from these problems: (i) the risk of network congestion due to the sent messages for updating data on resources and (ii) the risk of the churn effect if a large number of nodes want to update their data at the same time. These problems form big challenges in a large scale dynamic Grid environment. In this paper we propose a method of resource discovery and selection for large scale query optimization in a Grid environment. The resource discovery extends the P2P system Pastry. DHT are used to save only static data on resources. We retrieve the dynamic properties of these resources during the resource selection by a monitoring tool. First, the originality of our approach is to delay the monitoring of resources to the phase of resource selection. This strategy will avoid a global monitoring of the system that is often employed in the current resource discovery systems. Second, the method is executed in a decentralized way in order to help the database optimizer to better discover and select resources.