Adaptive Dataflow: A Database/Networking Convergence

Joseph M. Hellerstein
Professor, EECS Computer Science DivisionUC Berkeley


In recent years I have been noticing certain resonances between database and networking research, which I will try to amplify in this informal talk. Dataflow infrastructure is core to both communities, but there has been very little meeting of the minds on this topic up to now. A convergence of the networking and database approaches to dataflow should lead to a set of interesting technical interactions, spanning not only networking and databases, but also providing points of collaboration with our friends in Machine Learning and Control Theory. The application areas for this work are also lots of fun, and include sensor networks, peer-to-peer networks, and even the dusty old realm of the Internet. Commercialization opportunities will be left as an exercise for the audience.