Relating Requirements and Architectures: A Study of Data-Grids

Abstract

The requirements and architecture of any complex software system are highly interdependent. We have studied the relationship between these two concerns in several data-Grid systems. Data-Grids are characterized by an infrastructure that focuses on the coordinated management of, and access to distributed data resources. We survey current data-Grid projects to demonstrate that a set of general requirements for data-Grid systems can be identified. Architectural styles are a way of highlighting design and engineering similarities between software systems. We consider the styles that are exhibited by current data-Grids and use a lightweight methodology to analyze how these styles support general requirements. Our conclusions provide guidelines to assist the data-Grid developer in making informed architectural choices. In this paper, we present an extended case study of the relationship between the requirements and architecture of data-Grid systems. The architectures of current data-Grids can be shown to exhibit characteristics of various architectural styles. By analyzing how these styles support the core requirements of the domain, we can identify those styles that offer ‘best-fit’ and provide guidelines for the engineering of dataGrid systems. The relationship between requirements and architectures is not a concern unique to the Grid domain, but an area of active enquiry in the wider software engineering community. By means of this detailed study we hope to contribute to this discussion. Within an informal taxonomy of Grid systems, data-Grids are concerned with the generation of new information from distributed data repositories. DataGrids yield new information in various ways, by making available to scientists an unprecedented volume of useful data. They allow more rigorous statistical analysis, and enable the application of new data-mining techniques and the cross-correlation of sets of data that have not previously been compared. Data-Grids are characterized by an infrastructure that focuses on the coordinated management of distributed data resources and the provision of data access mechanisms.