3D VISUALIZATION OF EXTRA LARGE DATA ARRAYS BY MEANS OF SCIENTIFICVR ® VISUALIZER

Graphic presentation of the results in the form of color lay-outs or functional dependence diagrams makes it considerably easier to perceive new information and contributes to a more sophisticated and efficient analysis of complicated processes and phenomena. This becomes particularly important in the course of processing big 3D arrays that one has to deal with mainly when implementing computational simulation or processing experimental or analytic data. The existing visualization systems are characterized by a number of limitations. The main limitations are a maximal visualized data array size and bad 3D data distribution presentation. That's why the GDT Software Group has developed a new visualization system ScientificVR ® (SVR) offering numerous facilities and devoid of all the limitations typical for other systems. The main SVR advantages are: 3D voxel graphics, stereoscopic visualization, visualized data array size limit absence (this is of especial importance for data obtained from parallel computation systems) and on-the-fly visualization possibility. Voxel is a 3D analogue of pixel. The voxel is semitransparent, and its color is defined by the visualized function value. Voxels transparency allows the user to see the function distribution in the whole 3D area, not only in selected sections. Transparency can be adjusted to provide the most informative visualization. Stereoscopic visualization allows creating stereo images and animations. Stereo presentation is much more effective than usual and makes the data analysis process more comfortable. The visualized data array size limit absence is provided by means of a special data averaging procedure, which is to be done on parallel computation nodes by means of controlled downsampling procedure. As a result the maximal size of visualized data depends on the amount of RAM in a parallel computation system only. Required RAM size on the computation system rendering a visual presentation depends on the resolution of a resulting image only. Special programming strategy, including SDK, was developed to implement such behavior in an assumed, user-developed parallel application. This article is devoted to description of development ways of a new visualization system, working on various distributed computing systems with a random user application. The most prospective and efficient way to develop the new visualization system on the basis of a SVR visualization system, developed by the GDT Software Group is proposed.