Sangam: A Transformation Modeling Framework

Integration of multiple heterogeneous data sources continues to be a critical problem for many application domains and a challenge for researchers world-wide. One aspect of integration is the translation of schema and data across data model boundaries. Researchers in the past have looked at both customized algorithmic approaches as well as generic meta-modeling approaches as viable solutions. We now take the meta-modeling approach the next step forward. In this paper, we propose a flexible, extensible and re-usable transformation modeling framework which allows users to (1) model their transformations; (2) to choose from a set of possible execution strategies to translate the underlying schema and data; and (3) to access and re-use a library of transformation generators. In this paper, we present the core of our modeling framework a set of cross algebra operators that covers the class of linear transformations, and two different techniques of composing these operators into larger transformation expressions. We also present an evaluation strategy to execute the modeled transformation, and thereby transform the input schema and data into the target schema and data assuming that data model wrappers are provided for each data model. To show re-usability in our framework, we also present one transformation generator and show how the generator can produce a transformation model for any given input schema and data. The proposed framework has been implemented, and we give an overview of this prototype system.