Scientific Computing and Visualization: Linux Clusters and Tiled Display Walls

1
Scientific Computing and Visualization Linux
Clusters and Tiled Display Walls

• Glenn Bresnahan
• glenn_at_bu.edu

2
BU Scientific Computing Environment
Desktop Departmental Computing High-end Visualization HPC
1982 Ascii terminal VAX w/ Unix VAX/PDP-11 RasterTech Mainframe w/ vector
1989 Xterminal, Sun workstation Unix server (Sun, etc) SGI mp workstation CM-2
1996 SGI workstation SGI or Sun server SGI Onyx SGI Origin
2002 Intel-based PC (Win Linux) Unix server, Linux cluster Linux cluster IBM P3/P4, Linux cluster
3
SCV Linux Clusters

• IBM cluster
• 56 dual processor compute nodes (1 GHz P3)
• 24 graphics nodes (1.7 GHz P4)
• Myrinet interconnect
• GiB cluster
• Grid-in-a-Box test-bed
• Atlas test cluster
• Atlas/CMS test cluster

4
Linux on the Desktop

• Hardware
• Software
• Configuration
• Performance
• Applications and techniques

5
Linux in Visualization

• High-end visualization
• Big data, high-resolution, immersion
• Clusters and tiled displays
• Deep Vision Display Wall (stereoscopic, cluster
  driven, tiled display wall)
• Tiled displays and the Access Grid
• VR/VE on the AG
• http//scv.bu.edu/HiPArt/GoGrid/
• AG on cluster-based, tiled display walls

6
VR/AG Bridging
AG node
view renderer
video bridge
navigation
Multicast
DAFFIE
audio renderer
AG audio
AG node
viewer
7
Linux in HPC

• Compilers
• Batch systems
• Performance tuning
• Multiprocessing on clusters

8
Agenda - Tuesday

• 100-120 Welcome and Overview
• 120-220 Graphics hardware and software under
  Linux
• 220-240 Break 
• 240-350 Tiled display walls using Linux
  clusters - Building, using and maintaining a wall
• 350-420 Tiled display walls - Integration of
  spatial audio
• 420-450 Tiled display walls - Relation to the
  Access Grid and other systems
• 450-500 Wrap up

9
Agenda - Wednesday

• 1000-1200 Demonstration of Deep Vision Display
  Wall
• 1200-100 Lunch Break
• 100-200 Overview of scientific computing and
  Linux clusters
• 200-300 Visualization of scientific data under
  Linux - Techniques and data wrangling
• 300-315 Break
• 315-335 Visualization of scientific data -
  Domain-specific applications
• 335-405 Visualization of scientific data - Case
  histories
• 405-415 Wrap up

10
Agenda Thursday

• 900-1000 High Performance Computing on Linux
  Clusters
• 1000-1200 Introduction to MPI
• 1200-200 Lunch Break
• 200-315 Intermediate MPI
• 315-330 Break
• 330-500 Intermediate MPI (continued)
• Note Thursday sessions not available on AG