180 likes | 360 Views
GVis: Grid-enabled Interactive Visualization. State Key Laboratory. of CAD&CG Zhejiang University, Hangzhou 2005-12-06. Background (1). Scientific Visualization Widely utilized in many fields. Resource demanding (CPU, Memory, Graphic). Real-time User Interactive. ……. Materials.
E N D
GVis: Grid-enabled Interactive Visualization State Key Laboratory. of CAD&CG Zhejiang University, Hangzhou 2005-12-06
Background (1) • Scientific Visualization • Widely utilized in many fields. • Resource demanding (CPU, Memory, Graphic). • Real-time User Interactive. …… Materials Earth physics Medicine
Background (2) • Grid Computing • Integrating tremendous distributed resource on Internet. • Providing nontrivial services to end-users. • Conventional Grid System • Most dedicated to batch tasks with little user interaction. • Non-practical real time requirement.
Main problems • Grid resources management • Heterogeneous • Distributed • Dynamic • Throughput and Latency • Large data transporting over network • Rendering and Compositing
Our work • GVis: Grid-enabled interactive visualization system • Large scale resource utilizing. • Remote, parallel interactive visualization. • Multi-task visualization. • All Java based, Cross-platform.
GVis Architecture (1) • GVis (GVRE, GVVF, GVis Portal) • Grid middleware (Globus Toolkits 3.0) • JRE, OS (Windows 2000/XP, Linux)
GVis Architecture (2) • GVRE • Grid infrastructure for visualization application. • Based on Globus Tookits 3.0. • GVVF • Framework for local, remote, parallel interactive visualization. • Loosely coupling with GVRE. • GVis Portal • Interface between GVis user and GVis system. • Easy remote access to GVis system.
GVRE - GVis Runtime Environment • Information Management • Resource nodes information. • Visualization tasks information. • Implemented with OpenLDAP. • Resource Management • Updating resource node information. • Schedule and dispose request for resource usage. • Implemented as Grid Service. • Visualization Task Management • Task description. (XML file) • Task distribution, monitoring and migration. • Updating visualization task information. • Implemented as Grid Service.
GVVF - GVis Visualization Framework (1) • Components of GVVF • Visualization task scheduler – task partition. • Renderer – rendering. • Compositor – compositing partial results. • Data server – dynamic data loading. • Presenter – user interaction with visualization result.
GVis Visualization Framework (GVVF) (2) • Volume data visualization • Hardware based texture rendering. • 2D texture – faster, lower quality, more memory. • 3D texture – slower, higher quality, less memory. • Java OpenGL Binding (JOGL). • Convenient programming API. • Acceptable efficiency. • Partition and composition • Image space based partition. • Need whole data copy, not suitable for large dataset. • Object space based partition. • Need part of data, support large dataset.
GVVF - GVis Visualization Framework (3) • Optimization for User-Interactive • Directly Based on J2SE NIO Socket Channel. • Hardware based 2D/3D texture rendering. • Only useful image data is transported after rendering.
GVis Portal (1) • View GVis system information • Resource node information. • Visualization task information. • Customize visualization task • Submit and terminate visualization task • Interact with submitted visualization task
GVis Test and Result (1) • Large dataset visualization • Up to 3.75 GB. • 512 x 512 x 1024 x 15 • 30 nodes (rendering nodes, compositing node) • Cluster nodes connected with 1000Mbps Ethernet. • 1.0 fps