Leveraging open-source development in large-scale science data management systems

When completed, NASA's EOSDIS Core System (ECS) will be the world's largest Earth science data system, managing almost nine petabytes of data and disseminating more than two terabytes each day. The system's original design assumed that all science data archive, processing and dissemination would be done using high performance subsystems with complex distributed object interfaces between them. These interfaces made it difficult for others to extend the system without incurring the prohibitively high costs of supporting this infrastructure. Over the past three years, most of these interfaces have been replaced with greatly simplified script and file-based interfaces. NASA also has encouraged science and Data Center groups to modify and extend the system's core functionality. As these extensions began to emerge, it was apparent that new configuration management and system deployment methods would be needed to leverage each group's extensions across the ECS Data Centers. NASA has adapted several Open Source development techniques to address this need. This paper will describe how the ECS architecture and supporting development methods have evolved to support Open Source development concepts while at the same time satisfying ECS's challenging requirements. It also win describe how these changes have helped lower the system's overall costs and decrease the time it takes for new capabilities to become operational. We plan to build on the success of these initial changes to encourage additional EOSDIS user participation through many new roles: client and portal providers, data providers, algorithm providers, data processing centers, data service providers, distribution centers and data managers.