MAYA

For a decade now, researchers have talked about the emergence of “One Huge Computer“—an architecture that permits the Internet and its billions of attached computing devices to essentially function as a single, albeit massively distributed and decentralized, information system.

This provocative vision has motivated many of the most important developments in contemporary computer science. Mobile code, device discovery protocols, N-tier and peer-to-peer architectures, and many other advances were conceived in pursuit of this vision. We have a long way to go to achieve that vision, and many difficult problems remain to be solved, but the first steps on the path to the Universal Computer now seem fairly clear.

Far less clear, however, is the path toward another, equally provocative and challenging goal: the Universal Database. If we propose to build a computer that spans the planet, will we not also need an architecture for the persistent storage and organization of data that is capable of scaling—both in size and in space—to a similar degree? How can we simultaneously support the requirements for rapid, flexible local access, distributed implementations and administration, stability, evolvability, security, and all of the other imperatives that such a database would entail?

Such a system will not be built by assembling federations of relational databases interconnected via transactional links. The laws of physics suffice to guarantee that. Just as the needs of distributed, network-centric computing have demanded new design patterns, tools, and abstractions, creating the Universal Database will similarly require an architectural discontinuity from current practice. This paper describes the beginnings of such an architecture, and a component—the VIA Repository—designed in accordance with that architecture.