面向组合服务动态自适应的事务级主动伺机服务替换算法

An Active and Opportunistic Service Replacement Algorithm Orienting Transactional Composite Service Dynamic Adaptation

动态服务环境的各种异常随时会导致整个业务流程暂时无响应或服务中断,极大影响业务流程的可靠性.已有替换机制大多缺乏事务支持而适应性差,进而不能有效保证系统执行过程中事务服务的原子性和数据一致性,也无法保障替换过程要求的正确性、实时性和高效性.该文以"事务支持"为核心,充分考虑了服务间多关系以及Web服务的事务特性,提出一种事务级组合服务主动伺机替换算法.首先,该文给出了事务粒度的获取及替换范围识别算法.然后,提出事务级替换代价/收益QoS模型,将事务补偿代价与替换代价有机结合,保证以最少的代价实现服务替换.在此基础上,提出了全新的QoS驱动的事务级服务替换算法.为了提高替换的时效性,提出了早期预测模式挖掘算法,对运行的服务实时监控保证其识别失效服务的早期性.实验结果证明,该模型不仅保证了替换过程中业务流程的事务原子性和数据一致性,而且提高了系统的可靠性和时效性.

Various exceptions leading to the business process no response temporarily or service interruption are common in dynamic service environment,which reduce the reliability of business process greatly.Due to absent of transaction support,most existing replacement mechanisms are of poor adaptabilities.Therefore,the atomicity and the consistency of the transactional service can not be well assured during the system implementation process,and the correctness,the real-time performance and the efficiency of the replacement process can not be guaranteed either.This paper takes the ＂transaction support＂ as the core and proposes an active and opportunistic transactional composite service replacement algorithm,which fully takes into account the multiple relationships among Web services and the transactional characteristics of Web services.In this paper,the authors first present a mechanism to obtain the transactional granularity and a method to identify the cascading compensation range,then,propose a transactional replacement cost/benefit QoS model,which completes service replacement with the least cost by integrating transactional compensation service QoS and service replacement QoS.Further,a novel QoS-driven services replacement algorithm with the transactional compensation support is designed.In order to enhance the real-time performance,the authors introduce the early prediction pattern mining to identify the failed or unavailable services as earlier as possible by adopting a real-time monitoring to the running services.Experimental results show that the proposed model not only guarantees the atomicity and the data consistency of the process business in the replacement,but also improves the reliability and the real-time performance of the system.