Intervalo flexível para criação de checkpoints em simulação distribuída

Abstract

Simulation models with asynchronous components are subject to time violation. Checkpoints can be used to perform rollback operations and restore the simulation to a previous consistent state. In distributed models, components can independently checkpoints to reduce communication overhead, which can lead to the creation of useless checkpoints. Useless checkpoints are unwanted as they entail wasted processing and storage. In order to reduce useless checkpoints and the number of time spent in rollback, this paper presents an algorithm to calculate flexible intervals between checkpoints. The strategy was implemented in the Distributed Co-Simulation Backbone (DCB). The results showed that the implementation reduced the simulation rollback time by almost 10%, as well as a reduction of almost 24% in the number of useless checkpoints created.