Abstract: Traditional dynamic data race detection methods model activity causal relationship in a less precise way.Furthermore,mutual exclusion relationship is causal constraints that lead to various misreports or underreporting .To address this problem,a finer classification of causal relationships among Business Process Execution Language (BPEL) activities was proposed.A new logic clock-based method for finer identification of causal relationships was developed.Meanwhile,an accurate processing of mutually exclusive relationships among activities in BPEL programs was proposed by combining vector clocks and global mutex locks together.As a result,a more refined method for dynamic detection and prediction of data races in BPEL programs was proposed,which was able to ensure both a lower rate of data race false positives and a lower rate of data race negatives to a certain extent,but at the expense of more space and time.The advantages of the proposed method were verified in conjunction with a BPEL process example.

Key words: data race, concurrency defect, Web service composition, business process management, logic clock

摘要： 鉴于传统的数据竞争动态检测方法对因果关系的建模不够准确,且将互斥关系处理为因果约束,导致多种误报或漏报,对业务流程执行语言(BPEL)活动间的因果关系进行精细划分,提出一种基于新型逻辑时钟的因果关系精细化识别方法,并联合向量时钟和全局互斥锁对BPEL程序中活动间的互斥关系进行准确处理,由此提出一种更为精细的BPEL程序数据竞争动态检测和预测方法,该方法在一定程度上既能保证较低的数据竞争误报率,又能降低数据竞争的漏报率,但会牺牲更多空间和时间。最后通过BPEL流程实例验证了所提方法的优越性。

关键词: 数据竞争, 并发缺陷, Web服务组合, 业务过程管理, 逻辑时钟