We present and demonstrate capabilities of Multi-Feature Indexing Network (MUFIN), http://mufin.fi.muni.cz/. To achieve an independence of the similarity abstraction and the ability to process large collections of data, MUFIN is based on two basic paradigms: (1) the metric space model of similarity, and (2) the concept of structured Peer-to-Peer networks. In the demonstration we will show an interactive image retrieval system which indexes 50 million images - it is one or two orders of magnitude more than any other system designed to this purpose can do. Further more, the scalability of the system grows with constant complexity, which is very important for future internet technologies. To demonstrate the extensibility of the metric-based MUFIN approach we will show another prototype application: a face-recognition and retrieval system.