Improving Grid Nodes Coalitions by Using Reputation

In this work we deal with the problem of maximizing the QoS associated to each node of a Grid Federation, by modeling it as a problem of “Grid formation”. In the proposed model each node belonging to a computational Grid, is free to join with or leave a grid with the goal of improving its satisfaction. Contextually, each node is free to search other nodes to join with its grid or removing from the grid those nodes resulting ineffective. Software agents are employed to manage the node profiles; in our model we introduce a Grid agent having the role of managing the profile of the Grid. We introduce a distributed algorithm, called GF, to handle the node activity of joining to the grid, modelled as a matching problem. We present some experiments showing the effectiveness of our approach.