A study of storing large amount of inhomogeneous data in workflow management systems

Abstract. In workflow-driven Enterprise Resource Planning (ERP) systems a large variety of documents requires handling and storing of various descriptive data in a single storage facility. The number and type of attributes can vary among different kinds of documents. Storing such inhomogeneous data in a single database is difficult, as querying requires fast retrieval of data based on any present attribute.
In this paper the authors introduce three different approaches to this problem based on relational and document-oriented database systems. All the three solutions are compared by implementing and testing them with massive inhomogeneous data and by using a sophisticated cost model.
This problem is typical in ERP systems. Moreover, the solution can be generally applied to any domain using inhomogeneous data, like e-commerce systems, document warehouses and GIS systems.