Abstract: A new application disaster tolerance method was proposed to meet the disaster tolerance requirements of time-constrained longrunning applications. The method saved the middle running states of applications periodically and transfered these states to backup servers asynchronously. After disaster happened, application could resume from the latest checkpoint. Method of setting the checkpoint parameters was provided too. The prototype has verified the efficiency of the method in simulation environment.

Key words: information system, time constraint, Continuous Data Protection (CDP)

摘要： 针对带有时间约束的长时关键应用面临的容灾需求，提出一种基于检查点的应用容灾方法，在应用运行过程中定期保存应用的中间运行状态，并将中间状态异步传输到异地，灾难发生后应用无需重新运行，可自动从最近的检查点位置接力运行。给出自适应的检查点参数设定方法，并构造原型系统验证了该技术的有效性。

关键词: 信息系统, 时间约束, 持续性数据保护