Current Proceedings on Technology
Yazarlar: Mohammed Rebbah, Yahya Slimani, Abdelkader Benyettou, Boudjelal Meftah, Lionel Brunie
Konular:-
Anahtar Kelimeler:Large scale systems,Grid computing,Fault tolerance,Dynamic colored graph
Özet: The heart of the Grid is the ability to discover, allocate, and negotiate the use of system resources, such as a computer, network, or storage system. The Grid consists of geographically distributed cluster federations gathering thousands of nodes. At this scale, node and network failures are no longer exceptions, but a part of normal system behavior. Thus, grid applications must tolerate failures and their evaluation should take reaction to failures into account. We propose, in this paper, a decentralized model of fault tolerance based on new model of graph entitled dynamic colored graphs. The proposed protocol provides the ability to tolerate faults in all grid nodes, defining a threshold of Neighboring Collaborators for each node. After a coloring phase followed by a stabilization phase, we can classify the grid nodes into three categories: unstable, stable and hyperstable nodes. We demonstrated that the graph always converges to a stable state. We used the colors of the nodes to develop a new strategy for fault tolerance based on the colors assigned to nodes after the coloration and stabilization phases. From this model, we show through some experiments, the benefits of dynamic colored graphs to manage failures in grids.