Many new applications, for example Wikis, social networks, and distributed recommender systems, require the efficient integration of decentralized and heterogenous data sources at a large scale. In this paper, we present our vision of a universal storage for RDF-like triple data based on a structured P2P system and the universal relation model as its key enabling technologies to achieve flexibility, robustness, and efficiency for large-scale distributed data storage and query processing. We outline the steps we have already accomplished successfully and discuss a roadmap to achieve the final goal of a practical, light-weight implementation of our storage system.