The SCIJump Framework for Parallel an Distributed Scientific Computing

In recent years, software component technology has been a successful methodology for large-scale commercial software development. Component technology combines a set of frequently used functions in an easily reusable component and makes the implementation transparent to the users or other components. Developers create new software applications by connecting groups of components. Component technology is becoming increasingly popular for large-scale scientific computing to help tame software complexity resulting from coupling multiple disciplines, multiple scales, and/or multiple physical phenomena. In this chapter, we discuss our SCIJump Problem Solving Environment (PSE), that builds on its successful predecessor SCIRun and the DOE Common Component Architecture (CCA) scientific component model. SCIJump provides distributed computing, parallel components and the ability to combine components from several component models in a single application. These tools provide the ability to use a larger set of computing resources to solve a wider set of problems. For even larger applications that may require thousands of computing resources and tens of thousands of component instances, we present our prototype scalable distributed component framework technology called CCALoop. When the technology described in CCALoop matures, it will be included in SCIJump. SCIRun is a scientific PSE that allows interactive construction and steering of large-scale scientific computations [25, 27, 26, 18, 17, 19]. A scientific application is constructed by connecting computational elements (modules) to form a program (network), as shown in Figure 1.1. The program may contain several computational