Research On Petri Nets Based Web Service Composition Interaction Models And Their Appliacation Mechanisms

Currently, the Internet is not only information transmission infrastructure, but also information processing and services sharing infrastructure. Cloud computing is developing rapidly, the main concept of which is resources aggregation and on-demand services. So that future users do not care about how to buy the servers, software and solutions according to their business requirements, and only concern about how to obtain cloud services via the Internet to meet their own demands. Service-oriented computing (SOC), which provides network resources and services sharing with a uniform technical specifications, builds software application system by using services as basic units. Service-oriented architecture (SOA) has become a new generation of Web services infrastructure. In this framework, developers do not need to implement programs. They just write Web service composition scripts to build complex business applications.Software services and services collaboration are becoming a mainstream form of Internet applications. However, for the existence of malicious clients or abnormal partner services, service coordination may not be properly completed. For this reason, a qualitative analysis model is required to describe the interactions between all partner services as the purpose of runtime monitoring service composition. On the other hand, considering the relationship between service composition performance and Internet environmental factors such as network speed and reliability, it is needed to import quantitative Internet environmental factors to build a quantitative model based on the previous interaction model, in order to optimize service execution environment. On the basis of the latest research results related to this dissertation, it presents Web service composition interaction model, interaction behavior detection algorithm, runtime monitoring mechanism and model simulation based Internet environment optimization methods qualitatively and quantitatively. The key contributions of this dissertation include:Firstly, this dissertation proposes colored Petri net (CPN) based Web service composition interaction model WS-PIM for the purpose of qualitative analysis. Recently, the Business Process Execution Language (BPEL) serves as a service composition description language to describe service composition process. WS-PIM model focuses on collaborative interactions between partner services in the process of service composition. In the WS-PIM model, these synergistic interactions relationships are described by CPN. The model maps process status to the"token"distribution in the"place"of CPN, maps the changing of process to the"transition"of CPN, and maps the BPEL activities for process assembly to WS-PIM basic patterns. Then, the nested iteration and combination of WS-PIM basic patterns are implemented, by the way of"place"fusion. Several important properties of WS-PIM model are also provided. Experiments show that the model is more suitable than the similar models in describing the collaborative interactions of service composition process.Secondly, a temporal properties consistency based Web service composition interactions behavior detection algorithm is introduced. Temporal properties of interaction behaviors contain partial order, liveness, safety properties etc. In actual service composition process, the message is the only evidence of interaction behavior. These message sequences are representative of the actual interaction course; on the other hand, the changing course of service composition process status is showed by the transition firing sequences, which are obtained by analyzing WS-PIM model. The temporal properties reflected by the two sequences are consistent. Using a typical service composition process as an example, this dissertation gives temporal properties consistency based Web service composition interactions behavior detection algorithm. The algorithm captures the actual message sequences, and compares them with transition firing sequences to determine whether the execution of composite service is abnormal or not. Experiments show that the algorithm effectively detect the exception of service composition occurred during the process.Thirdly, the dissertation presents a runtime monitoring mechanism based on the Web service composition interaction behavior detection algorithm. This mechanism is composed of interaction model feature extractor, interaction behavior capturer and temporal property consistency checker. Interaction model feature extractor is capable of automatically transforming service composition script described by BPEL to WS-PIM model, and extracting the interaction behaviors and their temporal properties. Interaction behavior capturer is deployed in SOAP message engine, which can capture all the interaction messages related to monitored Web service composition process in real-time. These messages are forwarded to temporal property consistency checker. Temporal property consistency checker uses Web service composition interaction behavior detection algorithm to check the captured interaction behaviors. The core logic of checker is automatically generated by interaction model feature extractor.Finally, a generalized stochastic Petri nets (GSPN) based quantitative Web services interaction model WS-GIM is introduced. The model focuses on the relationship between the performance of Web service composition and network interaction environment. It concerns about the execution time of partners services and the speed of message transmission in the network and their failure rates, all of which directly affect the service composition system throughput performance. On the basis of WS-PIM model, WS-GIM model appends the service execution and message transmission two basic GSPN patterns. Under the Poisson distribution assumption, the mean parameters of individual services, such as service execution round trip time, service failure rate and process branch execution probability, are set in order to establish the quantitative WS-GIM model. By model simulation and statistical analysis, we acquire the average system throughput and failure rate of the whole service composition sample. The quantitative model can be used to evaluate the impact of interaction network environment on service composition throughput performance, so that it is applied to service composition performance optimization methods. At the same time, a composite service deployment plan selection method is also provided. By using the WS-GIM model prediction results, this method makes the use of multi-attribute utility theory to select preferable composite service deployment plan.