Towards a distributed execution framework for JUnit test cases

JUnit is the de-facto framework for creating and executing unit tests. This paper introduces a distributed execution framework for JUnit test cases called HadoopUnit, which is built upon Hadoop: an open-source platform for running applications that process vast amounts of data on large clusters built of commodity hardware. The primary motivation behind developing HadoopUnit was to test Hadoop production code using the Hadoop platform itself: existing approaches to testing the system were taking too long to run to completion, and were unable to provide feedback to the developers in a timely manner. Preliminary results suggest that HadoopUnit can reduce test execution time significantly: a 150-node cluster has produced a 30x improvement. HadoopUnit can be used by anyone for testing an application as long as they have access to a computing cluster with the Hadoop software installed and running.