Combining Optimism and Pessimism in a Grid Message Logging Protocol
Résumé
Fault tolerance is a key issue in grid systems. But few works have been done to provide transparent fault tolerance mechanisms for message passing applications executed in large scale grids. Optimistic and pessimistic message logging protocols are two well known solutions to provide fault tolerance for message passing applications. Although these two approaches seem to be completely opposite, we propose to combine them to create a grid fault tolerance protocol. This new protocol targets federation of clusters, a particular case of grid infrastructure.