Failures Handling Strategies Of Web Services Composition Base On Petri Nets

In recent years, SOC (Service-Oriented Computing) have been promoted by the rapidInternet technology development. A variety of Web services continue to be brought out on theInternet, but from a functional point of view, most Web services are function-limited, whichcannot meet the users’ increasingly complex business needs. To solve this dilemma, Webservices composition technology came into being. Web service composition technology haswidened the application of Web services technologies, digging out the genuine potential ofWeb services. Due to a strict mathematical foundation and straight-forward graphicalrepresentation, Petri nets, characterized by asynchrony and concurrency, are very suitable formodeling distributed systems, such as Web services.The petri-net based Web service composition execution engine is the realization of Petrinet theory for Web service composition. However, the complexity of Web servicescomposition execution engine causes uncertainty in the application system. Moreover,consider the situation where the application of a system fails; for a given input, it will notproduce the desired output, and the user’s needs cannot be met. At this point, knowing thefailure of the system is futile for the users, as they still cannot use it. Therefore, when asystem failure is discovered, it is of great significance to recover from a failure state in time.In order to restore the execution state of Petri-net based Web service compositionexecution engine, the traditional method is to reconstruct the corresponding Petri net.However, Web services for service composition are generally provided by different serviceproviders, and these network-distributed Web services are not free available, so it isunreasonable to discard the original Petri nets and remodel new ones.To tackle the failure of application systems based on Petri nets, this paper first defines theconcept of Atomic service execution failure, and meanwhile propose a discriminating methodfor this failure; the Petri-net based execution fault diagnosis of a Web service composition ispremised on discriminating method.Subsequently, this paper proposes a two-stage processing method for Petri-net based Webservice composition execution failure. The first level is Atom service-level troubleshootingmethods. Given traditional Petri nets, the concept of atomic services is defined, based on which fault handling policy for Web service composition execution is made. After that, bymapping the transition of Petri nets to Web services, the concept of similar Atomic services isproposed. Thus, the Petri-net transition replacement can be solved by combining with theservice feature short encoding algorithm used for failure service matching introduced inchapter4. The second level is subnet-level troubleshooting methods. It deals with the situationwhere Atomic services could not be found; by combining the service QoS dynamiccalculating method for replaceable candidate Web service ranking proposed in chapter3, thereachability of Petri nets can be recovered through a subnet substitution method.Finally, examples are used to demonstrate the feasibility of fault diagnosis and two-stagefailures processing algorithms proposed in this paper.