Dryad: A general-purpose distributed execution platform

Google Tech Talks November, 1 2007 ABSTRACT Web search has generated the need and economic support for a new class of data-intensive supercomputing applications. Several computing platforms have been created to support this need: the first described in the literature is Google's MapReduce. I will describe the architecture of the Dryad system developed at Microsoft Research, and explain some of our design choices. Dryad allows more general computations than MapReduce, and has consequently been used as a middleware abstraction on which higher-level programming models can be implemented. I will also briefly discuss some of these. Speaker: Michael Isard Michael Isard started out as a computer vision researcher, but has gradually been lured into systems research by his colleagues, first at DEC/Compaq SRC and now at Microsoft Research Silicon Valley. He was closely involved in the design and implementation of the first version of Microsoft's in-house search engine, and his systems research subsequently has concentrated on programming models for parallel and distributed computing.