Automatic Configuration for Running Models in the Cloud

I was really pleased to discover that the newsletter "International Science Grid This Week (iSGTW) referenced SCOOP in the lead article. But running one model in a distributed environment is only one step of a tall ladder. We don't have a world repository of grids and algorithms that we could assemble painlessly. For example, we should be able to ask for a numerical model that produces synthetic winds, storm surge or currents. And, the model should be described with enough information (aha Metadata) that will allow a computer program to run it. For example, information about the input required to run the model. A piece of software, with this information, could try to get the required input data  or discover another process or orchestration of processes  that could generated the required input.

When I start going up the ladder I imagine processes, inputs, parameters, phenomena, etc., getting categorized and related to each other (aha ontologies) to create automatic workflows, reusing grids, processes and real time observations available in the Web. The ladder contains multiples steps, some of which are being reinforced by MMI (metadata+ ontologies).  I only wish MMI could pull me up a little higher so I could see the world like if I where on a cloud.