Research On And Implementation Of Interactive Grid Visualization System GVis

The Grid is an important global information technology, which is arisen in the past few years. It provides the necessary functionalities and mechanisms for large-scale resources sharing and collaborative working between virtual organizations. Now it is used broadly in many areas of science and engineering. However, the majority of current grid systems are still batch-task oriented, which don't provides sufficient supports for interactive applications. In fact, there are a lot of interactive applications in science and engineering areas, i.e., scientific visualization. Due to its capability' of transforms large quantities of raw data into graphical images, visualization has been playing a very important role in scientific research, national defense and industry. Visualization is inherently computing and storage resource demanding, traditionally it is usually processed on high-performance computers or clusters. However, the scale of resource utilizing and service providing is not satisfied in the traditional visualization systems. Grid technology can enable visualization to utilize resources and provide services to a much wider extent, which favors a new research direction, interactive visualization grid. The main content of research on interactive visualization grid includes building grid to aggregate large-scale resources, creating visualization applications on the grid and providing interactive visualization services to end users. Among the current work on visualization grid, few have been done systematically, and most are based on existing parallel visualization systems, which are often platform dependent and not scalable. Besides, interactiveness is often neglected in a certain degree.In this dissertation, with all above points in mind, we propose and implement a java-based interactive visualization grid system GVis, in which real-time interactive visualization of large volume dataset is achieved by utilizing distributed and dynamic heterogeneous resources on the grid. Remote visualization service is also provided to end users through a grid portal in GVis system. Our focuses and achievements in this dissertation include:An extensible three-tier architecture for interactive visualization grid is proposed and implemented, which is composed of GVis runtime environment (GVRE), GVis visualization framework (GVVF) and GVis Portal. Each tier is responsible for a group of distinct functionalities. GVRE provides grid resources management, information management, visualization task management and visualization data management. GVVF is responsible for creating concrete remote and parallel visualization applications on the grid. And the grid portal GVisPortal is the interface for end users to access and interact with the services of GVis system. Low coupling and high extensibility is achieved with this three-tier architecture upon which GVis is implemented.GVRE is responsible for providing basic functionalities and mechanisms for interactive visualization applications to execute effectively, stably and reliably on the grid. Grid middleware Globus Toolkit 3 does not support for building interactive visualization grid directly, so the function of Globus Toolkit 3 is carefully selected and necessary extensions to Globus are built in GVRE. First, based on the grid service management mechanisms in Globus Toolkit 3, two kinds of grid services, LocalResourceManageGridService and TaskProxyGridService, are created and deployed to accomplish resource management and task management on the grid. Second, a customized Index Server is built on LDAP other than using the Index Service of Globus Toolkit 3directly. Last, task monitoring and migration mechanism, as an extension to Globus Toolkit 3, is implemented to enhance the fault-tolerance ability of G Vis system.To utilize the heterogeneous grid resource effectively, GVVF is totally implemented in Java language, which provides remote and parallel visualization applications on the grid. Hardware texture based rendering algorithm and two kinds of parallel rendering methods, which are image space based paralleling and objects space based paralleling, are implemented in GVVF to support large dataset visualization. Due to fully object-oriented design, extensibility is achieved in GVVF, new rendering algorithms and parallel rendering methods can be added to this framework easily.GVisPortal provides the interface for end users to access and interact with visualization service of G Vis system. Instead of traditional web-based grid portal, GVisPortal is implemented as a Java application based grid portal, which provides strong supports for end users to use the system and consume the visualization service interactively. GVisPortal is a cross-platform portal which can be deployed and executed on many important platforms, such as Windows and Linux.GVis is a quite comprehensive interactive visualization grid system with a lot of work been done. It is totally implemented in Java, which makes it a cross-platform system and really practical.