Building a High-resolution Scalable Visualization Wall

Abstract

High-resolution and scalable displays are increasingly being used for interactive 3D graphics applications, including large-scale data visualization, immersive virtual environments, and scalable 3D rendering subsystems. A display wall is a unique environment by combining a high-resolution display surface and a parallel computing cluster. Most display wall systems today are built with high-end graphics machines. We are investigating how systems that use only inexpensive commodity components of a PC cluster are build. Rocks is an excellent commodity-based cluster management toolkit, which centers around a Linux distribution based on Red Hat. Additionally, Rocks allows end-users to add their own software via a mechanism called Rolls. Viz is one of the Rolls provided in Rocks to build tiled display visualization clusters. In this thesis, we describe the architecture, major components and render algorithms of this Viz visualization cluster in detail. We also show how to build, configure, and implement visualization programs with several approaches. Our experiment validates the framework and principles of the cluster as well as performance improvement in terms of service availability, load balancing brought about using a sort algorithm.