State of NERSC Address

NERSC Accomplishments in 1996

http://www.nersc.gov/research/whitepaper/whitepaper.html

NERSC Mission Statement

Provide reliable, high quality, state of the art computing resources and client support in a timely manner--independent of client location--while wisely advancing the state of computational and computer science.

Major Milestones

Feb 96	Four NERSC employees in Berkeley
April 96	First J90 in Berkeley
May 96	C90 and storage move
May 96	User Services move
May 96	PDSF arrives
June 96	Staff move complete - ERSUG meeting
Aug 96	new ERCAP process complete
Aug 96	Cray 2's turned off
Sep 96	T3E and two additional J90s installed
Oct 96	Grand Opening
Nov 96	Supercomputing '96
Dec 96	NERSC budget "White Paper"
Jan 97	T3E "inching" toward acceptance

FY 1996 - Transition to new NERSC

FY 1996 Accomplishments	FY 1997 Major Goals
physical move to LBNL installation of J90s and T3E recruited more than 50 new staff members 20% budget cut vs. LLNL	transition to new operational principles and models integration of T3E, J90 cluster, new mass storage build team, new culture, intellectual infrastructure build new programs/projects

The Intellectual Home of NERSC

Computational Science Competency

Computer Science Competency

Hardware (evaluation)	RISC architectures hierarchical memory systems interconnection networks parallel I/O emerging architectures systems performance evaluation
Software	software engineering (00, functional programs, distributed computing) languages and tools (HPF, HPC, spl;it C, PVM, MPI, ... , debugger, performance monitoring tools, libraries)
CS Applications	scientific database load balancing and scheduling visualization and graphics algorithms

The Challenge

• Computer technology continues to evolve rapidly
• The new NERSC must be more proactive in adopting computer science technology
• NERSC can only afford a small amount of research FTE's

Future Technology Mission

Visualization Group

• Work closely with NERSC researchers
• Provide generic visualization tools for use in applications
• Develop strategies and tools to allow remote NERSC users to utilize our visualization resources
• Maintain a variety of visualization software packages on different platforms
• Educate and consult with researchers on the various visualization tools available

NERSC Enhancements and Initiatives

• Grand Challenge Proposals - FY1997
  • Expect 6 GCAs to work closely with NERSC staff
  • Integrated teams with user services, visualization, scientific computing, and future technology staff
  • No incremental funding for supporting GCAs
• LDRD
  • Berkeley Lab approved computational LDRDs in collaboration with NERSC/CSD
  • Total funding is more than $3M
  • About 20 postdocs and students will work on these projects
  • Additional 2-3 FTEs for NERSC
  Computational Science Program with:
  • Joint appointments for computational science training
  • Training classes developed for NERSC users
  • Seminar series broadcast via MBone
  • Results from small group projects posted on WWW
  All will directly benefit the NERSC community at large
• Bay Area
  Berkeley provides an intellectually rich environment, many informal interactions, e.g., with:
  • ICSI, Berkeley (International Computer Science Institute)
    J. Feldman (pSather parallelization tool)
  • MSRI, Berkeley (Mathematical Sciences Research Institute)
    D. Hoffman (minimal surfaces)
  • Corporate headquarters for three of the five major computer vendors
  • NERSC has benefitted through close collaboration with SGI/Cray and Sun
• Benchmarking
• Architecture Evaluation/Tera
• NERSC Parallel Algorithm Prototypes

Summary

National/International Impact

• NERSC made top 25 list in Nov. 1996
• NERSC tutorial and booth at SC '96
• NERSC is ahead of intellectual competition because of:
  • re-engineering the center
  • focus on computational science
  • focus on computer science
• . . . and all this with a reduced budget