Interactive Visualization of Large-Scale Architectural Models over the Grid

Transcription

1 Interactive Visualization of Large-Scale Architectural Models over the Grid XU Shuhong, HENG Chye Kiang, SUBRAMANIAM Ganesan, HO Quoc Thuan, KHOO Boon Tat

2 Agenda Motivation Objective A Grid-Enabled Visualization Infrastructure Multi-Stage Data Compression Prototype System Deployment Summary

3 Motivation Virtual reconstruction of Tang period Chang an city

4

5

6

7 How to enable remote users with PCs to explore this splendid ancient city?

8 Objective To explore Grid and visualization techniques and to formulate an appropriate framework for interactive remote rendering of large-scale architectural models

9 A Grid-Enabled Visualization Infrastructure The promise of Grid computing: a transparent, interconnected fabric to link data sources, computing (visualization) resources, and users into widely distributed virtual organizations (Shalf and Bethel 2003) Relevant works (next slide) In summary, Grid-based visualization, especially for large-scale data, is far from mature till today

10 Remote & collaborative visualization System/approach Vizserver vcollab Brief Description Limitations Others Rendering is done at server site Sending data to each collaborator SGI supercomputer only No data protection Commercial Commercial Access Grid (AG) Open source group-togroup teleconferencing tool Not designed for rendering Several systems have been deployed over AG Grid-enabled visualization gviz project Grid extension of IRIS Explorer & pv3 Developed for these two software tools Research Grid Visualization Kernel(GVK) Aimed at a universal Grid visualization service No user interaction Research RAVE A resource aware visualization system Not designed for large-scale data Research

11 Infrastructure of A-VizGrid (architectural visualization Grid) Grid resources are used to do computation and handle data. On top of a resource, a visualization or data service can be deployed. These services are registered into a centralized information service so that they can be located by a visualization planner. A visualization or data service runs within its execution service container which deals with communications over the Grid using Grid-compliant protocols

12 User request display input parameters user interactions Portal Server image compression and transfer using UDP Visualization Service frame buffer rendering data receiver resource info resource info request Visualization Planner Information Service resource info reply data transfer using GripFTP resource info Data Service data source data processing data transfer launch of applications launch of applications Infrastructure of A-VizGrid execution container execution container

13 Considerations of A-VizGrid - Sharing rendering burden over the Grid (1) make use of remote rendering facilities (2) protect designers ownership of 3D models - Large data sets through limited network bandwidth (1) distribute rendering tasks to a limited number of powerful rendering facilities connected by high-speed networks (2) use data/image compression techniques - Launching applications through execution service containers

14 Multi-Stage Data Compression Pre-processing stage - divide the city into districts - simplify 3D models Rendering stage - view-dependent culling - level of details (LOD) - data paging Remote display stage - image compression and decompression

15 Dividing the city into districts Simplifying models into different resolutions

16 1:32 compressed by the SGI Vizserver algorithm 1:128 compressed by a new wavelet algorithm

17 Prototype System Development Based on the A-VizGrid infrastructure, an experimental prototype system is being developed Deployment of the prototype system - the Chang an city models are kept at National University of Singapore - available rendering facilities include a SGI Onyx2 with four graphics pipes at the Institute of High Performance Computing (IHPC), a Linux cluster with 10 AMD64 Opteron CPUs and 5 Nvidia Quadro FX3000G graphics cards at IHPC, and a SGI Onyx3000 at Nanyang Technological University

18 - rendering software tools include OpenGL Performer with enhanced culling and view-dependent continuous LOD algorithms, VRCO CAVELib, and Chromium, etc OGSA-based Grid services - information service - visualization planner - execution service container (using Java, completed) A simple demonstration of distributed rendering using Grid services and execution container

19

20 Summary To enable remote users with normal personal computers to interactively visualize large-scale architectural models, a Grid-based visualization infrastructure A-VizGrid is proposed The underlying Grid services are developed according to the OGSA standard A multi-stage data compression approach is used to improve rendering speed.

21 Thank You