Workflow history management

Abstract

A workflow history manager maintains the information essential for workflow monitoring and data mining as well as for recovery and authorization purposes.Certain characteristics of workflow systems like the necessity to run these systems on heterogeneous, autonomous and distributed environments and the nature of data, prevent history management in workflows to be handled by the classical data management techniques like distributed DBMSs. We further demonstrate that multi-database query processing techniques are also not appropriate for the problem at hand.In this paper, we describe history management, i.e., the structure of the history and querying of the history, in a fully distributed workflow architecture realized in conformance with Object Management Architecture (OMA) of OMG. By fully distributed architecture we mean that the scheduler of the workflow system is distributed and in accordance with this, the history objects related with activities are stored on data repositories (like DBMSs, files) available at the sites involved. We describe the structure of the history objects determined according to the nature of the data and the processing needs, and the possible query processing strategies on these objects using the Object Query Service of OMG. We then present the comparison of these strategies according to a cost model developed.