Limitação do espaço de armazenamento de checkpoints em simulação distribuída otimista

Abstract

Distributed simulation models that have asynchronous components are subject to time backtracking caused by time violation. Checkpoints are used for restoring consistent states prior to time violation. Throughout the simulation the components can generate non-useful checkpoints in case of throwback. In large simulations these useless checkpoints take up more and more memory, and the lack of limited storage of these checkpoints can lead the simulation to a collapse due to lack of memory. The objective of this work is to present some techniques of limiting the storage space of checkpoints and some strategies to indicate when the checkpoints can be removed without prejudice to the simulation. The techniques and strategies will be implemented in Distributed Co-simulation Backbone (DCB) in order to perform experiments to evaluate the performance and usefulness of the results according to the proposed objective.