Presenter: Chandrajit Bajaj

Abstract:

At the Center of Computational Visualization (CCV) of the University of Texas at Austin, we have a commodity PC cluster consisting of 130 Compaq 750 workstations (nodes). Each node has one 800 Mhz Pentium III processor, 2 UltraScsi-3 hard drives (9.1 gig and 18.2 gig), a low-end graphics card, a CD ROM, and 256MB RAM. All machines are connected through 100Mb/s Ethernet, while portions of the cluster are also connected by Servernet II and Gigabit Ethernet. The cluster is in a four-row configuration, with each row having 32 Compaq SP750s. Each row has a name for reference; the names are Red, Green, Blue, and Alpha. The Alpha row has both low-end graphics cards and high-end Nvidia and ATI graphics cards.

We have implemented a suite of parallel volumetric image processing and visualization applications on this cluster. All of those applications have also been implemented as CORBA services that can be accessed remotely from client machines, using our volume ROVER client. In this talk I shall present several performance and scalability results of these remote visualization services. Details shall also be given of a visualization portal under development that enables and facilitates, interactive and batch use of CCV's high performance parallel services.