Evaluation of rollback points on multiprocessor systems
Kiyomitsu Yoneda, Takashi Matsubara, Yoshiaki Koga · Systems and Computers in Japan · 2000
Rollback recovery is a well-known technique to recover the system from failures for fault-tolerant properties. There may be a domino effect when this method is applied to multiprocessor systems. In this paper, we discuss the method that restarts from consistent checkpoints without message loss in the whole system, when failure occurs under the assumption that the processors are connected with reliable communication. Furthermore, we present a method to obtain checkpoints by using temporal logic to roll back rapidly in each processor from the current renewal checkpoints and communication. © 2000 Scripta Technica, Syst Comp Jpn, 31(11): 24–30, 2000