Abstract - In the last decade, as a consequence of the multidisciplinary and data-driven character of modern science, researchers grew a strong demand for collecting, integrating and combining information objects from multiple research-oriented data sources. The objective is to improve the way research results are shared, discovered, and re-used across community boundaries. Interoperability issues of data and technologies, scarcity of computational resources, and highly evolving requirements typically represent an obstacle to practitioners constructing the “aggregative infrastructures” capable of performing such an integration. This paper presents the architectural principles and the services of the D-NET Software Toolkit, a framework where developers find the tools for constructing and operating aggregative infrastructures in a cost-effective way as instances of service-oriented data infrastructures. In D-NET developers can select from a variety of data management services, can configure them to handle data according to given data models, and can combine them into autonomic workflows to obtain personalized aggregative infrastructures.