DART: Distributed Automated Regression Testing for Large-Scale Network Applications

This paper presents DART, a framework for distributed automated regression testing of large-scale network applications. DART provides programmers writing distributed applications with a set of primitives for writing distributed tests and a runtime that executes distributed tests in a fast and efficient manner over a network of nodes. It provides a programming environment, scripted execution of multi-node commands, fault injection, and performance anomaly injection. We have implemented a prototype implementation of DART that implements a useful subset of the DART architecture and is targeted at the Emulab network emulation environment. Our prototype is functional, fast, and is currently being used to test the correctness, robustness, and performance of PIER, a distributed relational query processor.