Table of Contents
Future Directions in Computer and Systems Architecture for Scientific Computing
Outline of This Talk
The National Academic Usage of Supercomputers Continues To Grow Exponentially
PPT Slide
Can PetaFLOPS Transition us from Computing as a means for Explanation to Computing as a mode for Discovery?
Best Prepared Applications Communities
Less Prepared Applications Communities
PPT Slide
PPT Slide
PPT Slide
“Simplified” Application View of an HTMT “Slice”
IBM’s Blue Gene Machine
IBM Blue Gene .. Software and Applications
Chiba City
Chiba City System Details
PPT Slide
Example: 320-host Clos topology of 16-port switches
Evolution of Cluster Services
Alarming Trends of Systems Software Availability from Computing Hardware Vendors
HPC Community is Leveraging Commodity Hardware
Targeted Hardware Systems
Possible Software Stack for HPC
Node Issues
Some Cluster Node Options Today
Intel Pentium III
IA-32 Willamette Demo’d @ 1.5 GHz
AMD Athlon v Pentium III
AMD K7/Athlon (www.amd.com)
Compaq Alpha 21264/21364
Intel IA-64
Intel IA-64 Building Blocks
Future IBM PowerPC (Power4)
IBM’s Power4
Conclusions
Possible Extreme Linux Systems in 2010 (conservative)
Some Cluster Node Options Tomorrow
Compaq Itsy Pocket Computer
IPic – Match Head Size Web Server
Are We Destined to Compute on Toys?
Each vector unit has enough parallelism to complete a vertex operation (19 mul-adds + 1 divide) every seven cycles.
Playstation2 Specs
PPT Slide
IRAM Vision: Intelligent PDA
ISTORE Hardware Vision
Berkeley’s Vector IRAM
Block Diagram
V-IRAM Design Overview
VIRAM-1 Floorplan
Vector Lanes
V-IRAM Prototype Summary
Petaflops from V-IRAM?
The Computational Grid ? Moving From Local to Global Access to Computing and Information Resources
The Emerging Concept of a National Scale Information Power Grid
The Grid Links People with Distributed Resources on a National Scale
Examples of Grids Testbeds
The Vision for the Grid
The Grid from a Services View
Midwest Networked CAVE and ImmersaDesk Sites
Access Grid Nodes
CorridorOne ? Distance Visualization
Acknowledgements
Many Thanks to Our Research Collaborators
More Thanks Owed For:
The FL Group at Argonne/Chicago is:
PPT Slide
Questions?
Questions?
Questions?
Questions?
Questions?
Questions?
Questions?
Questions?
Questions?
Partnerships for Advanced Computational Infrastructure Partner and Leading Edge Sites
Universities With Projects Using NSF Supercomputers
University High Speed Networks NSF vBNS and Internet2 Abilene
Prototyping America’s 21st Century Information Infrastructure
The Next Step is the Integration of International Research Networks
A Telecommunications Trend
PPT Slide
The Vision for the Grid
The Grid from a Services View
PPT Slide
Globus-Enabled Support for Priority Traffic on the Internet
Active Spaces ? Exploring Future Workspace Environments
Organizational and Social Trends
Active Spaces ? Workspace Environments
StationOne
PPT Slide
Proposed: Totally Active Work Space (TAWS)
Active Spaces ? As Spaces
Active Spaces ? As Grid Nodes
Midwest Networked CAVE and ImmersaDesk Sites
PPT Slide
PPT Slide
Advanced Display Environments
PPT Slide
PPT Slide
Directions for Active Space Development
Active Mural ? Beyond the MegaPixel
ActiveMural, a Tiled Display Wall
PPT Slide
The Active Mural Lab with Access Grid Tech
Corridor One ? Attacking the Distance Visualization and Resource Problem
CorridorOne ? Distance Visualization
PPT Slide
Our Belief and Conclusion
ActiveMural, ?Mural: Price Performance
The ?Mural
PPT Slide
Projector Alignment/ Image Blending
PPT Slide
PPT Slide
PPT Slide
PPT Slide
Research Directions for ActiveMural, ?Mural
Access Grid ? Integrating Group to Group Collaboration and Visualization
Group-to-Group Interaction is Different
Access Grid Nodes
Components of an Access Grid Node
PPT Slide
PPT Slide
PPT Slide
The Access Grid is More Than Teleconferencing
The Five Stages of Joint Work
Group-to-Group Requirements
The Docking Concept for Access Grid
The Illinois “True Grid” Dark Fiber Infrastructure
Proposed Terabit/s Midwest Network Testbed