The Vortex Approach to Integration and Coordination of Workflows

While many organizations use workflow management systems, few tools or techniques have been developed to support the interaction and coordination of workflows residing in different organizations. We distinguish two related aspects of such coordination, namely transactional (concerned with ensuring atomicity and durability of specified groups of actions) and semantic (concerned with the flow of information between, and intended meaning of, groups of actions performed by different workflows). This position paper identifies several issues that arise in connection with the semantic aspect of workflow coordination. The paper goes on to illustrate how a new declarative workflow paradigm called Vortex (see [HLS+99]) can be used to overcome these issues. It is crucial to use a clearly defined framework and language for specifying workflow interaction. The framework must address the heterogeneity between workflow systems and schemas, and between data representations. The language must support data translations and restructurings, coordination of the input and output of existing workflows, and semantic enrichments and extensions of existing workflows. Furthermore, it is essential that the framework and language facilitate formal analysis of workflow interactions, such as whether data flow dependencies between data in different workflows can lead to deadlocks. Vortex is a new programming paradigm for specific application domains, including in particular, workflow. The focus of Vortex programs is on information gathering, decision making, and launching and monitoring of external tasks. The Vortex language is declarative: programmers specify the conditions under which decisions and tasks should be performed, but ∗Part of work was done while visiting the Bell Labs. and part of work was supported by NSF grants IRI-9411330 and IRI-9700370.