A language and toolkit for the specification, execution and monitoring of dependable distributed applications

Abstract

This thesis addresses the problem of specifying the composition of distributed applications out of existing applications, possibly legacy ones. With the automation of business processes on the increase, more and more applications of this kind are being constructed. The resulting applications can be quite complex, usually long-lived and are executed in a heterogeneous environment. In a distributed environment, long-lived activities need support for fault tolerance and dynamic reconfiguration. Indeed, it is likely that the environment where they are run will change (nodes may fail, services may be moved elsewhere or withdrawn) during their execution and the specification will have to be modified. There is also a need for modularity, scalability and openness. However, most of the existing systems only consider part of these requirements. A new area of research, called workflow management has been trying to address these issues. This work first looks at what needs to be addressed to support the specification and execution of these new applications in a heterogeneous, distributed environment. A co- ordination language (scripting language) is developed that fulfils the requirements of specifying the composition and inter-dependencies of distributed applications with the properties of dynamic reconfiguration, fault tolerance, modularity, scalability and openness. The architecture of the overall workflow system and its implementation are then presented. The system has been implemented as a set of CORBA services and the execution environment is built using a transactional workflow management system. Next, the thesis describes the design of a toolkit to specify, execute and monitor distributed applications. The design of the co-ordination language and the toolkit represents the main contribution of the thesis.