Integrated Data, Message, and Process Recovery for Failure Masking in Web Services

Modern Web Services applications encompass multiple distributed interacting components, possibly including millions of lines of code written in different programming languages. With this complexity, some bugs often remain undetected despite extensive testing procedures, and occasionally cause transient system failures. Incorrect failure handling in applications often leads to incomplete or to unintentional request executions. A family of recovery protocols called interaction contracts provides a generic solution to this problem by means of system-integrated data, process, and message recovery for multi-tier applications. It is able to mask failures, and allows programmers to concentrate on the application logic, thus speeding up the development process. This thesis consists of two major parts. The first part formally specifies the interaction contracts using the state-and-activity chart language. Moreover, it presents a formal specification of a concrete Web Service that makes use of interaction contracts, and contains no other error-handling actions. The formal specifications undergo verification where crucial safety and liveness properties expressed in temporal logics are mathematically proved by means of model checking. In particular, it is shown that each end-user request is executed exactly once. The second part of the thesis demonstrates the viability of the interaction framework in a real world system. More specifically, a cascadable Web Service platform, EOS, is built based on widely used components, Microsoft Internet Explorer and PHP application server, with interaction contracts integrated into them