Cluster Computing for Web-Scale Data Processing

In this paper we present the design of a modem course in cluster computing and large-scale data processing. The defining differences between this and previously published designs are its focus on processing very large data sets and its use of Hadoop, an open source Java-based implementation of MapReduce and the Google File System as the platform for programming exercises. Hadoop proved to be a key element for successfully implementing structured lab activities and independent design projects. Through this course, offered at the University of Washington in 2007, we imparted new skills on our students, improving their ability to design systems capable of solving web-scale problems.