This project will address the needs of researchers involved in solving large nonlinear problems, which exhibit multiple solution paths, or large-scale minimization problems with multiple local minima. It will develop a library of visualization and computational steering tools to address path following problems efficiently on Beowulf computer clusters. It will also provide tools for optimization of the cluster communication. It will produce tools that are integrated with the CUMULVS steering and visualization environment. The software libraries will be open source and provided to the community of researchers and users.
The PI's are developing an environment for visualization and interactive computational steering on clusters of PCs. The visualization and steering front-end being implemented will be capable of interacting using a client-server model with a computational engine, which could also reside on a parallel computer, or a distributed network of workstations in addition to the cluster of PCs which is the main target architecture. The target application of this environment will be 3-dimensional liquid crystal material in a confined geometry. The model used is based on the Landau-de Gennes theory. In particular, one is interested in the changes in the degree of order and orientation of the liquid crystals as the temperature is varied. Knowledge about the properties of liquid crystals is of importance, for example, for the design of liquid displays. The target application is characteristic of a class of problems which would benefit significantly from computational steering, since it has multiple solutions with different symmetry structures and exhibits bifurcation or turning-point behavior for certain values of the determining parameters. Hence this is a path following or continuation problem.
As the workstation clusters become powerful and affordable, efforts are underway to interconnect a set of such clusters with a high bandwidth interconnection (such as ATM) to provide Network-Based Computing (NBC) environments. Leading examples of applications include collaborative computing, interactive and real-time visualization, multimedia, and geographically-distributed databases. In this research, we plan to take on new challenges in providing fast and scalable communication support for SMP-based network-based computing systems. This includes developing schemes for balancing computation, communication, and synchronization on such systems by taking the three-level hierarchy of communication into account; and studying application-level performance benefits of the proposed communication support.
