追求千兆級的運算：全球主導行動

Abstract

This document is about In Pursuit of Petascale Computing: Initiatives Around the World

Table of Contents

• Table of Contents
• IDC Opinion
• In This Study
• Situation Overview
  • Petascale Initiatives Around the World
    • DARPA HPCS Program: Cray Cascade Program (Peter Ungaro)
    • DARPA HPCS Program: IBM PERCS Program (Rama Govindaraju)
    • A Petascale Initiative in China (Zeng Yu, Dawning)
    • European Petascale Planning (Michael Resch, HLRS/Stuttgart)
    • Los Alamos National Laboratory: The Roadrunner System (John Morrison)
    • National Energy Research Scientific Computing Center (Horst Simon)
    • National Science Foundation Initiative (Jim Kasdorf, Pittsburgh Supercomputing Center)
    • Oak Ridge National Laboratory (Doug Kothe)
    • Petascale Computer Projects in Japan (Makoto Taiji, Riken)
  • Perspectives from HPC Experts
    • Petascale Algorithms (David Bailey, Lawrence Berkeley National Laboratory)
      • Petascale System Trends
      • Concurrency and Amdahl' s Law
      • The Challenges of 100,000-Way Concurrency
      • High-Resolution Climate Modeling on NERSC-3
    • What Software Architecture Will Cope with 100,000 Processors? (John Gurd, Manchester University)
      • Analysis
      • Rule of Thumb
    • Petascale or Petaflops? (Paul Muzio, Chair, HPC User Forum)
    • The View from Berkeley (David Patterson, University of California-Berkeley)
      • How to Measure Success
      • Why Should This Work
      • Summary
    • A Productive Petaflop is a Hard Sell (David Probst, Concordia University - Montreal, Canada)
    • Are Programming Model Changes Needed for Petaflops? (Panel Session, Manchester University)
      • Andrew Jones, University of Manchester
      • Ben Ralston, AWE
      • Paul Muzio, Chair, HPC User Forum
      • Ian Reid, NAG
      • Stephen Wheat, Intel
    • We Have 10 Years to Construct the Petaflop Ecosystem (Martin Walker, HP)
    • Thermals are the Key (Stephen Wheat, Intel)
  • IDC Survey Highlights Related to Petascale Computing
  • Table: Typical Number of Processors That ISV Applications Use for Single Jobs
  • Table: Ability of Application to Scale to Thousands of Processors
  • Table: Ability of Application to Scale to Tens of Thousands of Processors
  • Table: Key Factors Needed for ISVs to Improve Applications
  • Table: Use for a Petascale Computer
  • Table: Willingness to Run Heterogeneous Problems
  • Table: Size of Problems to Solve
• Future Outlook
• Essential Guidance
  • Guidance for HPC Users/Buyers
  • Guidance for HPC Vendors
• Learn More
  • Related Research
  • Synopsis