How computers have broken science (and how we can fix it)
Science relies on the ability to explain natural occurrences in a reproducible way. This reproducibility is a fundamental part of the scientific method, and is essential for ensuring that the experimental design is correct and that any subjective bias has been eliminated. A high level of transparency, and the ability to independently verify research, is critical for establishing the dependability of the results. As extremely powerful computer systems have become cheaper and more accessible, scientific research is increasingly being performed using modeling and simulation (M&S) rather than (or in conjunction with) experimentation in real life. In some fields, such as engineering, M&S is now the accepted method for research and development. Unfortunately, the widespread adoption of M&S is leading to a reduction in the level of transparency necessary for acceptable scientific research. In addition to requiring specialized hardware and software, a particular simulation might require a complex workflow involving an ensemble of applications, hand-crafted input data, and a significant amount of post-processing and analysis to produce usable results. Current tools make it very difficult to capture the exact environment and precise workflow necessary to reproduce complex simulations such as these.
The field of simulation lifecycle management (SLM) is emerging as a unified approach for the control of simulations and associated test data in a manner that promotes transparency and collaboration. Although there have been some efforts to converge on a set of tools and methods for SLM across scientific disciplines, these approaches are still ad-hoc and difficult to scale. Eclipse is an ideal platform for addressing the needs of SLM, and projects such as the Eclipse Integrated Computational Environment (ICE) and the Data Analysis Workbench (DAWNSci) are attempting to solve many of these problems.
This talk will provide an introduction to SLM, and show how it can address many of the issues facing scientist undertaking M&S-based research. We will then take a dive into some of the specific features of these projects to show how Eclipse is positioned to become the cornerstone platform for managing M&S across the entire scientific research community.