公开(公告)号：KR100875431B1

公开(公告)日：2008-12-22

申请号：KR1020070036808

申请日：2007-04-16

Applicant: 아주대학교산학협력단

Inventor： 조위덕 ,  김성수 ,  최창열 ,  장요철

IPC: G06F17/00 ,  G06F21/22

Abstract: A fault prediction model based aggressive checkpointing method is provided to reduce the return time and non-operation time, thereby increasing the seamless supply of application programs without applying any excessive overhead to a system. A fault prediction model based aggressive checkpointing method includes the steps of setting a checkpoint interval of a system, determining the number of entire checkpoints by dividing a failure prediction point supplied by a predetermined fault prediction model by the checkpoints, computing the number of aggressive checkpoints carried out in a fault section by excluding the number of normal checkpoints to be carried out in a normal stage from the number of entire checkpoints, storing system snapshots by carrying out the checkpointing in the normal stage and the fault section respectively by a predetermined interval according to the computed number of the checkpoint, and rollbacking the system to a point nearest to the point of failure by using the system snapshot at the time of a system error.

Abstract translation: 提供了基于主动检查点的故障预测模型，以减少返回时间和非操作时间，从而增加应用程序的无缝供应，而不会对系统施加任何额外开销。 基于故障预测模型的主动检查点设置方法包括以下步骤：设置系统的检查点间隔;通过将由预定故障预测模型提供的故障预测点除以检查点来确定整个检查点的数目;计算携带的检查点的数量 通过从整个检查点的数目中排除要在正常阶段执行的正常检查点的数目，在正常阶段和故障区段中分别按照预定间隔执行检查点存储系统快照 计算出的检查点的数量，以及在系统错误时使用系统快照将系统回滚到最接近故障点的点。

公开(公告)号：KR1020080093469A

公开(公告)日：2008-10-22

申请号：KR1020070036808

申请日：2007-04-16

Applicant: 아주대학교산학협력단

Inventor： 조위덕 ,  김성수 ,  최창열 ,  장요철

IPC: G06F17/00 ,  G06F21/22

CPC classification number: G06F11/0766 ,  G06F11/0793

Abstract: A fault prediction model based aggressive checkpointing method is provided to reduce the return time and non-operation time, thereby increasing the seamless supply of application programs without applying any excessive overhead to a system. A fault prediction model based aggressive checkpointing method includes the steps of setting a checkpoint interval of a system, determining the number of entire checkpoints by dividing a failure prediction point supplied by a predetermined fault prediction model by the checkpoints, computing the number of aggressive checkpoints carried out in a fault section by excluding the number of normal checkpoints to be carried out in a normal stage from the number of entire checkpoints, storing system snapshots by carrying out the checkpointing in the normal stage and the fault section respectively by a predetermined interval according to the computed number of the checkpoint, and rollbacking the system to a point nearest to the point of failure by using the system snapshot at the time of a system error.

Abstract translation: 提供了一种基于攻击性检查点的方法，以减少返回时间和非操作时间，从而增加应用程序的无缝供应，而不会对系统造成任何过多的开销。 基于攻击性检查点的方法的故障预测模型包括以下步骤：设置系统的检查点间隔，通过将由预定故障预测模型提供的故障预测点除以检查点来确定整个检查点的数量，计算携带的攻击性检查点的数量 通过从整个检查点的数量排除要在正常阶段中执行的正常检查点的数量的故障部分，通过按照预定的间隔分别在正常阶段和故障部分执行检查点来存储系统快照 计算出的检查点号，并在系统错误时使用系统快照将系统回滚到最接近故障点的点。