In this paper, Ozaki, Dohi, and Kaio discuss the problem of checkpoint placement, focusing on dependability in a centralized system. The dependability measures are steady-state system availability and expected reward per unit time in the steady state. The paper is very interesting because, when a failure occurs, the system recovers by rolling back to a previous checkpoint; the checkpoint is chosen depending on what failed and the timing of the failure.

The case studies explore two models. In the first model, the system stops during checkpointing; in the second, faults can occur during checkpointing. In the second model, the conventional algorithms used to obtain steady-state availability cannot be used; therefore, the paper presents new algorithms that can be used to solve problems for aperiodic checkpointing protocols.

The numerical evaluation and the experiments presented in this paper will enable students and scientists to improve their knowledge of checkpoints and provide new points of view for an area that has been studied for a long time. The numerical computation algorithms investigate quantitatively the optimal checkpointing algorithms and methods, and show their potential for practical application. The paper has a good bibliography review that provides a survey for people interested in this area.