摘要： 现有的Web事务工作流恢复算法通常未考虑用户的个性需求，也没有考虑恢复时的连锁恢复情况。为此，定义全局事务工作流模式和执行事务模式2种失效恢复粒度，并提出基于这2种粒度的失效恢复算法。通过细粒度的执行事务模式进行事务实例级的层次式递归恢复，采用粗粒度的全局事务工作流模式进行模型级的恢复，并用来满足客户的个性需求。该算法能够在保证用户需求的基础上动态确定补偿终止点，可有效减小补偿域，并能从全局视野上寻找补偿路径，实现连锁失效恢复。

关键词: 事务工作流, 失效恢复, 补偿, 双粒度, 连锁失效恢复

Abstract: The Web service transactional workflow failure recovery algorithm always does not take the individual requirement of user and the chain failure recovery into consideration. This paper defines two failure recoveries grained the global transactional workflows model and executive transaction model, and proposes failure recovery algorithms based on the two grained. The transactional instance level hierarchical failure recovery is performed under fine-grained executive transaction model. The model-level recovery is performed under coarse-grained global transactional workflow to maintain the individual requirement of client. The algorithm determines the end compensation point dynamically and compensation costs are reduced, while in the process of recovery, the algorithm can search the compensation route under the global viewpoint and implement the chain failure recovery.

Key words: transactional workflow, failure recovery, compensation, double granularity, chain failure recovery

中图分类号: 

• TP393