Research On Fault-Tolerance Of Web Services Driven Business Processes

As an emerging technology for distributed computing,web services have received much attention in both academy and industry recently.It is promising as it enables the interaction between and integration of heterogeneous software applications that are developed by different technologies.In order to simplify the business logic and improve service reusability,a single web service only provides limited functionality.It is therefore necessary to compose several existing services to form a value added service(i.e.,business process) to meet the complex business requirements in the real world.As web services in a business process are loosely coupled,and as they are heterogeneous and autonomous,failures are expected to occur frequently during the execution of the business process.How to reduce failure occurrence and recover business processes from runtime failures therefore becomes a significant problem in the area of web services.To address the above problem,we propose several fault tolerant techniques for business processes in this thesis.Our techniques take into account the heterogeneity and autonomy of web services,and thus can improve the reliability of business processes in the context of web services.The main research works and contributions of this thesis are described as follows:1.We propose a rule-based approach to handle exceptions in business processes.Our approach separates the normal business logic from the exception handling logic,and thus simplifies the development and maintenance of business processes.Exception handling logic is specified by ECA rules which build upon a set of exception handling patterns that have high level semantics.At runtime,these ECA rules are first automatically transformed into standard WS-BPEL codes and then integrated into normal business logic.Experiments show that our approach is feasible.2.We propose a constraint-aware compensation mechanism.Given a business process,our mechanism can ensure its execution atomicity by the least cost. We define three-level compensatability,and devise an algorithm to judge the compensatability of a business process.We also analyze the relationship between compensation cost and service scheduling,and propose two algorithms to find the optimal schedule which makes the least compensation cost.Experiments show that our approach can reduce the compensation cost while ensure the execution atomicity,and thus improve the QoS of business processes.3.We propose a profit-driven service selection algorithm to assist business processes to strike a balance between service reliability and service usage fee.We propose a business model for business processes,and analyze the relationship between the profit of a business process and the QoS of its partners.According to the analysis,we give a selection algorithm which chooses services with high reliability but keeps the service usage fee low at the same time.We conduct various experiments and the results show our selection algorithm performs well.4.We propose an approach to predict the QoS of business processes.For each kind of process structure,we give a set of reduction rules for transaction property.By using these reduction rules,we can predict the transaction property of a business process based on the transaction properties of its partners.In addition,we investigate the effects of fault tolerant structures on different QoS attributes,which in turn give us guidelines to use these fault tolerant structures in business processes.