Towards Personalized Multimedia Dial-Up Services

Recent advances in networking coupled with rapid advances in storage technologies are making it feasible to design a variety of sophisticated multimedia services, in which clients will not only access mundane services, such as television, radio, and telephones, but they will also have refreshing freedom to choose, procure, and personalize these services to suit their individual needs. In this paper, we present a service architecture that describes how these multimedia services are likely to evolve from a symbiosis of different enterprises: Network providers, who own high-speed networks; Storage providers, who control access to multimedia servers; and Service providers such as publishing houses, libraries, news distributors, and entertainment houses. Towards realizing such services, we present a system architecture that addresses the two most distinguishing requirements of media: intra-media continuity and inter-media synchronization. Defining a sequenceof continuously recorded media units (e.g., audio samples or video frames) as a Media Strand, we devise methods by which a storage provider can organize the physical storage of media strands on disk so as to guarantee their continuous retrieval, at real-time rates. In order to embody the temporal relationships among media strands, we define a Multimedia Rope as a collection of media strands intertwined by synchronization information. We indicate how multimedia presentations, such as lectures, movies, news, and even personalized video channels can all be represented as multimedia ropes. For synchronous playback of a multimedia rope, we present a technique by which a storage provider can determine the temporal relationships among the media strands of that rope, at the time of recording, and can later enforce these relationships, at the time of playback. At the Multimedia Laboratory at UCSD, we have developed a prototype multimedia dial-up service in an environment of multimedia workstations and video peripherals. We share our initial experiences with this prototype and outline our future plans. This work was supported by the IBM Corporation, the NSF, UVC Corp., Starlight Networks Inc., Compaq Corp., and the University of California MICRO program.