An important aspect of eScience is the development of new methods and tools to support scientists to enhance the ways they conduct research and to optimize the route to new scientific discovery. The route from data to information to knowledge and insight can and should take optimal advantage of modern ICT facilities and e-infrastructures, but often requires specialist experience. Ideally, researchers should be engaged in their scientific challenges rather than with ICT, however supportive.
The goal of eStep is to develop tools, interfaces, and libraries to deal with and extract information from large amounts of (distributed) data, requiring large computing infrastructures, high-speed networks, and high-resolution visualization equipment. Moreover, in many cases data and results, as well as compute kernels and full scientific workflows, are made sharable among multiple collaborating parties.
Avoiding 'reinventing the wheel'
As part of our strategy we explicitly try to avoid 'reinventing the wheel' even for domain-specific eScience solutions. To this end, eStep follows a layered and modular approach. At the lowest layer, system-level libraries are developed (or adopted) and integrated, following five themes:
The low-level libraries specifically aim at hiding the particular idiosyncrasies of accessing, and making optimal use of, the underlying hardware and middleware infrastructures. eStep tools categorized under the 'Computing' theme, for example, allow easy access to, and concurrent use of, a large variety of computing resources, thus supporting the Jungle Computing paradigm.
Ideally, researchers should be engaged in their scientific challenges rather than with ICT
Maximizing software re-use
The highest layer offers generic and domain-specific solutions. Each of the tools at this layer either serve as a complete solution for a domain-specific problem, or as a domain-independent 'skeleton' that can be used for various scientific problems with very similar properties. An example skeleton represents a multi-model/multi-kernel simulation environment that applies to both the domains of computational astrophysics and climate modeling, and possibly more. An important aim of the eStep project is to have the high-level tools be built on top of the low-level libraries as much as possible, thus maximizing software re-use.
Image: Martijn van Dam