The Scientific Computing Group

Mission

The Scientific Computing Group will facilitate development of scientific applications than run on NERSC capability platforms and promote optimal use of NERSC resources.

Responsibilities

• Collaborate with strategic users to port and develop scientific applications
• Evaluate, integrate, and create new hardware and software technologies
• Evaluate, integrate, and create new new numerical and non-numerical algorithms
• Aid in procurement of new capability systems
• Serve as internal NERSC quality control and user advocate
• Bridge gap between new technology and research scientists
• Bridge gap between research scientists and staff

Early tasks focus on preparing facility, staff, and users for T3E

Now - T3E Delivery:

• Learn T3E architecture and performance
• Educate NERSC users and staff; prepare users (User Services)
• Plan for T3E arrival (Large Systems)
• Evaluate ERCAP proposals for MPP allocations (ERSUG)
• Develop "terms of engagement" for collaborations
• Arrange early access for staff and users (CRI, other Vendors)
• Staff to 3+ group leader

Later, focus on demonstrating viability of T3E to meet needs of ER researchers

T3E Delivery - Optional Upgrade:

• Collaborate with small number of strategic projects (Scientific Visualization/User Services)
• General development and porting support (User Services)
• ensure CRI meets production status requirements for optional upgrade
• Educate NERSC users (User Services)
• Contingency planning/follow on procurement planning
• Staff to 10+ group leader

Future tasks are more pathfinding in nature

Optional Upgrade -- ?

• Development of advanced capability applications that stress systems
  • Hardware
  • Software
  • Numerical Algorithms
• Integration of research into production environment
• Identify areas for research in response to unmet user needs
• Set direction for high-end computing

NPAPs (NERSC Parallel Algorithm Prototypes) will aid users in developing parallel applications

• Repository of parallel algorithm prototypes
  • computational techniques
  • tradeoffs
  • performance
  • program development tools
• Range from kernels to state-of-the-art computational techniques used as building blocks

NPAPs will enable NERSC to impact a larger number of users than a one-on-one approach

• Ease burden of getting started on parallel systems
• Reduce lag time between development of new algorithms and introduction into applications
• use to develop parallel applications and save in program development costs

Group composed of 10 scientists with broad-base knowledge of scientific computing on capability platforms

• Discipline-specific knowledge (Office of Fusion Energy, High Energy and Nuclear Physics, Basic Energy Sciences, Health & Environmental Research, Computational & Technology Research)
• Computational Science knowledge (ODEs, PDEs, linear and non-linear solvers, image processing, particle methods, etc.)
• Parallel Computing (architecture, software systems, performance, etc.)
• Computer Science (scientific databases, object-oriented design and development, distributed computing, etc.)

Members of the Scientific Computing Group

Tammy Welcome, Group Leader
Sisira Weeratunga
Adrian Wong
Sherry Li
Youngbae Kim
Chris Ding
Osni Marques
Andrew Canning
+ 2 openings