基于服务关系统计的多粒度服务组合方法

Multi-scale service composition approach based on statistics of service relationship

传统工作流方法进行服务组合存在两个问题,服务组合无法自动生成和复杂服务无法重用,基于服务关系统计的多粒度服务组合方法(MSWC)有效解决了这两个问题。MSWC包括服务关系统计学习、服务粗分和服务细分三个部分,分别用于进行服务关系统计计算、服务分组和服务映射。通过服务关系统计学习计算了有逻辑和无逻辑的服务关联度;服务粗分将服务需求分解为无逻辑关系的组;而服务细分通过逻辑概率将服务分组映射成系统中已有的复杂服务,将这些复杂服务串接起来,即完成服务组合。因为三个步骤均是自动执行,因此MSWC是自动组合方法,而服务分组最终映射为系统已有的复杂服务,因此是一个多粒度服务组合方法。最后通过实验结果分析表明,该算法能够很好地适应网络上Web Service快速增长的情况,并且具备较好的服务组合性能。

Based on statistical methods, Muhi-Scale Web service Composition （MSWC） approach optimizes traditional workflow method in two aspects： automatic orchestration and complex service reuse. There are three steps in MSWC, service learning, coarse-grained service decomposition and fine-grained service decomposition. Service learning majors for calculating service probabilities, including logic and non-logic. Inspired by divide-and-conquer, services in the request list are divided into several groups in coarse-grained service decomposition, service logics are determined and service groups are mapped into existed atom and complex services in fine-grained service decomposition. And service composition is achieved by cascading all groups. For three steps are all automatic and existed complex services are used, MSWC is an automatic and complex service reuse approach for service composition.