Abstract
The MapReduce system built on commodity hardware is the de facto standard for large scale data analysis. In this talk, we will describe its relationship to PRAMs and BSP, present new algorithms tailored specifically for this paradigm and conclude with some open questions.