Abstract: Recovery-Oriented Computing(ROC) improves system availability by repairing the system as soon as possible, instead of avoiding failure. This paper studies ROC techniques, gives its application in cluster system, proposes the method of Recursive Restartability(RR) based on node group and Undo recovery model based on Checkpoint to improve system availability. It evaluates the improvement effect of the methods.

Key words: availability, Recovery-Oriented Computing(ROC), recursive restartability, Undo model, cluster

摘要： 针对面向恢复计算(ROC)技术致力于在故障发生后使系统尽快恢复，从而提高系统可用性，而非从根本上避免故障发生的特点，对面向恢复的相关技术进行研究，给出ROC技术在集群系统中的应用，提出基于节点组的递归重启方法和基于Checkpoint的Undo恢复模型，用以提高集群系统的可用性，并对2种方法的改善效果进行评估。

关键词: 可用性, 面向恢复计算, 递归重启, Undo模型, 集群

CLC Number: 

• N945