This document discusses processing large graphs in Hadoop. It describes Google's Pregel framework, which takes a vertex-centric approach to iterative graph processing. Pregel computations occur over multiple supersteps, with vertices sending messages to each other between steps. The document also covers Apache Giraph, an open-source implementation of Pregel built on Hadoop. Giraph allows graph processing jobs to leverage Hadoop features like HDFS and MapReduce. An example shortest path algorithm is provided to illustrate Pregel's message passing model.