A high‐speed recovery method for use in distributed computer systems
Yuzuru Maya, Toshiyuki Kinoshita, Isao Yoshino, Yoshinori Tokunaga, Satoshi Takemura · Systems and Computers in Japan · 1992
Abstract A system is proposed for high‐speed switching from the on‐line unit to the back‐up unit in hot‐standby systems for high reliability. In a hot‐standby system it is necessary to transfer periodically the memory contents of the on‐line unit as checkpoint data to the memory of the back‐up unit so that the back‐up unit can take over in case of a system fault. The proposed method guarantees that there is no file update between consecutive checkpoints by transferring checkpoint data whenever a file is updated. Because of that, the journal for file recovery is not necessary, thus making the downtime short. The downtime and overhead of this method are compared quantitatively with those of the conventional method. The results show that in this proposed method the overhead is within the allowable range, the downtime is about 2 s, and the system downtime cannot be recognized by users.