Software Development and Testing: Approach and Challenges in a distributed HEP Collaboration

In developing the ATLAS [1] Trigger and Data Acquisition (TDAQ) software, the team is applying the iterative waterfall model, evolutionary process management, formal software inspection, and lightweight review techniques. The long preparation phase, with a geographically widespread development team required that the standard techniques be adapted to this HEP environment. The testing process is receiving special attention. Unit tests and check targets in nightly project builds form the basis for the subsequent software project release testing. The integrated software is then being run on computing farms that give further opportunites for gaining experience, fault finding, and acquiring ideas for improvement. Dedicated tests on a farm of up to 1000 nodes address the large-scale aspect of the project. Integration test activities on the experimental site include the special purpose-built event readout hardware. Deployment in detector commissioning starts the countdown towards running the final ATLAS experiment. These activities aim at both understanding and completing the complex system, and help in forming a team whose members have a variety of expertise, working cultures, and professional backgrounds.