Dynamic Environment Driven Computational Science and its Terascale Data: Keeping the Human in the Loop
Computational science, the third pillar of scientific enterprise, is essentialt o solving of complex problems ranging from the environment to systems biology to weather forecasting. Computational science, traditionally the bastion of large parallel applications, today faces application domains wanting to respond to the real world immediately with complex model execution. In the LEAD project we are building a service-oriented architecture (SOA) that supports dynamic weather forecasting, that is, weather forecasting that can be carried out on demand in response to a sudden regional weather event. Reacting to the environment in real time requires a handful of key components that will be topic of this talk: a workflow system capable of dynamic adaptation, dynamic filtering and screening of large volumes of data, and a data management subsystem that provides automated meta data generation and rich query searching across highly heterogeneous data sources. In this talk we identify the requirements, challenges, and lessons learned from building a framework for dynamic weather forecasting.