A domain specific model for generating etl workflows from business intents

Extract-Transform-Load (ETL) tools have provided organizations with the ability to build and maintain workflows (consisting of graphs of data transformation tasks) that can process the flood of digital data. Currently, however, the specification of ETL workflows is largely manual, human time intensive, and error prone. As these workflows become increasingly complex, the users that build and maintain them must retain an increasing amount of knowledge specific to how to produce solutions to business objectives using their domain's ETL workflow system. A program that can reduce the human time and expertise required to define such workflows, producing accurate ETL solutions with fewer errors would therefore be valuable. This dissertation presents a means to automate the specification of ETL workflows using a domain-specific modeling language. To provide such a solution, the knowledge relevant to the construction of ETL workflows for the operations and objectives of a given domain is identified and captured. The approach provides a rich model of ETL workflow capable of representing such knowledge. This knowledge representation is leveraged by a domain-specific modeling language which maps declarative statements into workflow requirements. Users are then provided with the ability to assertionally express the "intents" that describe a desired ETL solution at a high-level of abstraction, from which procedural workflows satisfying the intent specification are automatically generated using a planner.