A Study On A Grid-Enabled Environment For Engineering And Scientific Computation

Computational grid environments usually include a variety of geographic distributed heterogeneous resources with different architectures, different operating systems, different configurations, different vendors and different availabilities. The resources are autonomous, and reside in different administrative domains with different security control policies. The usability and availability of the resources may dynamically change during the period of sharing. Computational grid systems not only need to overcome resources heterogeneity and accommodate resources autonomy, but also support efficient resources discovery, dynamic resources allocation, effective tasks monitor and control.With enlarging of the scale and increasing of precision in multidisciplinary application simulations, more and more large-scale data are produced. Problem resolving with supercomputers is costly. Scientists are searching more inexpensive and more powerful computational power for their numerical analysis and visualization in scientific and engineering computing. Nowadays, computational grid technologies are utilized to meet such a demand.This dissertation provides a MASSIVE grid system. It is a multidisciplinary applications oriented simulation and visualization environment, especially for numerical simulations in CSM (Computational Solid Mechanics) and CFD (Computational Fluid Dynamics). It is intended to provide an integrated toolkit with numeric analysis, parallel computation and visualization for scientific and engineering researchers.1. Research on the MASSIVE architecture.Based on Globus Toolkit 2.2/2.4, MASSIVE is composed of three layers, construction layer, services pool layer and application layer. It introduces visual grid service concepts, and includes visual grid information service, visual file transfer services, visual resources allocation service, visual task scheduling service and visual task monitor service.2. Research on visual information services and the EISA architecture.We have developed a visual information services system MASSIVE-MDS. It provides a customized resources information discovery service with an application programming interface.EISA (Extensible Information Service Architecture) consists of three layers, the registration and deployment layer, normalization and organization layer, and information collection layer. With consideration of virtual organization participator's autonomous control of local resources and more detailed resource information requirements by grid users, self-defined dynamic information discovery programs can be deployed and extended easily in the EISA architecture. EISA improves extensibility of grid information services and efficiency of resources utilization.3. Research on efficient resources scheduling algorithms and resources broking policies.To improve the accuracy of task scheduling and resources utilization, MASSIVE...