This document discusses scaling genetic algorithms using MapReduce. It motivates using MapReduce for genetic algorithms applied to large-scale problems as it provides a simple, scalable programming model compared to MPI. It outlines the genetic algorithm process and how it can be modeled in MapReduce by partitioning the population across nodes, using mappers and reducers to perform operations like selection and crossover in parallel, and optimizing representation. Experimental results show it can scale genetic algorithms to problems with 100 million variables on a commodity cluster.