The complexity of modern science and engineering is daunting in terms of the complexity of analysis and processing, and volumes of distributed, heterogeneous data that must be managed. The cost of creating the software systems to support and manage these large-scale endeavors is significant, and historically most tools have been custom built and tailored specifically to particular scientific and engineering problems.
This talk will describe the architecture-driven approach we have successfully employed at PNNL to build core technologies that are broadly applicable across scientific and engineering disciplines. We rely heavily on leveraging robust, widely adopted open-source technologies and integrating them in novel ways through flexible software architectures. This creates functional, flexible systems at considerably lower costs than custom-designed technologies. We illustrate this approach by describing our Velo knowledge-management system, which highlights our ,build systems, not codeŠ mantra.