Data-Flow Anti-patterns: Discovering Data-Flow Errors in Workflows

摘要

Despite the abundance of analysis techniques to discover control-flow errors in workflow designs, there is hardly any support for data-flow verification. Most techniques simply abstract from data, while data dependencies can be the source of all kinds of errors. This paper focuses on the discovery of data-flow errors in workflows. We present an analysis approach that uses so-called "anti-patterns" expressed in terms of a temporal logic. Typical errors include accessing a data element that is not yet available or updating a data element while it may be read in a parallel branch. Since the anti-patterns are expressed in terms of temporal logic, the well-known, stable, adaptable, and effective model-checking techniques can be used to discover data-flow errors. Moreover, our approach enables a seamless integration of control flow and data-flow verification.