Research On And Implementation Of Grid-enabled Java-based Interactive Visualization System GVis

Visualization transforms large quantities of raw data into graphical representations and has become an indispensable tool in scientific computing, national defense and industry. However, the process of visualization is inherently so compute and data intensive that it usually needs high-end hardware and parallel techniques to make it applicable and practicable. With the fast advances of computer hardware and network from the late 1990's, the performance of graphics hardware and network communication have improved dramatically. As a result, visualization has run in a more interactive and distributed mode. Grid, as an enabling technology to Internet wide distributed computing, enables visualization to utilize resource over and provide services to a wider extent of Internet and favors a new research direction, i.e., grid-enabled visualization. Grid-enabled visualization provides resource sharing and task management capabilities to interactive parallel, remote and collaborative visualization. Current research on grid-enabled visualization is still in its preliminary stage, few work have been down systematically and little attention has been paid to resource heterogeneity, task management and application interactiveness. This dissertation is a relatively systematic research with these points in mind. A grid-enabled Java-based interactive visualization architecture and system GVis is proposed and implemented with which interactive remote and parallel visualization on dynamic and heterogeneous resources is possible. Our focuses and achievements in this dissertation include:An extensible three-tier architecture for grid-enabled visualization is proposed, which is composed of a grid supporting tier, a visualization tier and a grid portal. Each tier is responsible for a group of distinct functionalities. The grid supporting tier provides resource management and task management to visualization tasks. The visualization tier is responsible for concrete visualization and parallel visualization implementation. And the grid portal is the interface for users to access the grid-enabled system. Low coupling and high extensibility is achieved with this three-tier architecture upon which GVis is implemented.Grid-enabled visualization systems utilize heterogeneous resources and accept visualization tasks dynamically. However, most of current parallel visualization systems can not provide dynamic resource management and multiuser multi-task support. Moreover, grid middleware Globus does not supportinteractive visualization tasks directly. Hence, the grid supporting tier is introduced and the function of Globus is carefully selected to provide high performance resource and task management.As most of current parallel visualization systems are designed for homogeneous platforms only to achieve high performance, they can hardly handle the heterogeneity nature of grid resources. Java, as a cross-platform language, has the advantage of built-in platform independence support. Tough suffering from performance penalty, Java has been improving its efficiency steadily during the past ten years. By using Java as the underlying platform for the visualization tier and with an object-oriented design, platform independence, system simplicity and extensibility are achieved. Hardware textures based volume rendering is used as the default method for volume rendering due to its high efficiency.Grid users access grid through grid portals, which usually has a web-based interface and is more effective for non-interactive tasks. The interactiveness in visualization tasks can hardly be satisfied by such web-based grid portals. Instead of traditional web-based grid portals we proposed a Java application based grid portal design, providing strong supports to interactiveness in remote parallel visualization for large datasets.The work of this dissertation is a relatively systematic research and implementation of grid-enabled visualization system. GVis provides platform independence, resource sharing and task management to remote and parallel visualization and is a good platform for future research and extensions such as grid-enabled collaborative visualization.