In this talk we present an R package for handling huge hierarchical statistical data. Hierarchical means that entities can contain datasets describing or referencing subentities on several levels. By huge we understand data far larger than the main memory. This data is stored in a MySQL database allowing simultaneous access from multiple users, however no SQL knowledge is needed to access this data. The framework is flexible enough to store any kind of data without modifying the database structure. Appropriate data views can be read and updated directly in the form of data frames. Our approach handles references, access rights and graph structures in the data. It uses an object oriented framework supporting data classes and inheritance.