A Research Of Service Availability And Related Exception Handling For Service-Oriented Applications

With the evolvement of business environment, more and more attention has been gained inthe cross-organization and personalized application integration domain. Traditionaltechnologies cannot offer enough solid support due to lack of features to deal with the newproblems(such as heterogeneous and flexibility in the dynamic and open environment. As apromising distributed computing paradigm for network-based application integration,Service-Oriented Computing (SOC) can serve as advanced enabling technology forloose-coupled application integration. However, services in SOC are autonomous, which makethe application integration process unstable. On second thoughts, the problem is regarded asthe exception caused by the unavailability of services, which is furthermore represented by theuncertain connection between the application being built and the services being used. Thisthesis brings up a novel approach to handling these exceptions by means of fully utilizing theunderlying services. It proposes the corresponding model, patterns, supporting mechanism, andpresents a prototype system supporting the approach. The main research work andcontributions include:1. Based on the analysis of the source of service unavailability, we adopt the idea ofECA(Event-Condition-Action) to make full use of the underlying available services tocome up with our exception handling approach: VERSATILE. According to the testapplication and comparison with other related work, we conclude that VERSATILE canenhance the utilization of available services and consistency of the application. Also, itsoperational goal is to involve the business user to make it adaptable to the personalizedapplication development.2. Mechanism of exception prediction and avoidance. Using event based trackingmechanism, changes of service availability can be predicted, and the idea of "servicereservation" is adopted to enhance the consistency of application executing. Theadvantages are: system handles the changes of service availability to make it transparent tobusiness user;prediction and handling process have little effect on the service automationand system loosely coupling.3. Exceptions can be handled through adapting the connection between application andservices by business user involvement. Based on category theory, we defined theconnection adapting model to fetch up the fact that there is no common theory to supporthow to build the connection between application and services.4. To handle the exception, we define the business user configurable adaptation mechanismand algorithms. By using the algorithms, the side effect caused by the unavailable servicewill be eliminated. Then a new matching process will be started to search matchingservices according to the configuration predefined by business user.Four patterns are brought forward to instruct how to handle exception by involvingbusiness user. Those patterns are based on the possible interactions between business users andsystem. They can maximally utilize the ability of business user during handling exceptions.To use above approach, we designed and implemented an exception handling prototypesystem. Using the prototype system and the experimental requirements from the FLAME2008 project as a test bed, we proved the feasibility and validity of our approach.