Center for Enabling Distributed Science

1
Center for EnablingDistributed Science
A Proposal for a DOE ScIDAC-2 Center for Enabling
Technology
2
DOEs SciDAC Program

• DOE Office of Science
• Office of Advanced Scientific Computing Research
  (OASCR)
• Other DOE offices
• SciDAC-1 2001-2005
• Large-scale simulation
• Collaboratories program (e.g., PPDG)
• SciDAC-2 2006-2010
• Science Applications
• Enabling Technologies

3
SciDAC-2, contd.

• Create a new generation of Scientific Simulation
  Codes that take full advantage of the
  extraordinary computing capabilities of terascale
  computers
• Create the Mathematical and Computing Systems
  Software to enable the Scientific Simulation
  Codes to effectively and efficiently use
  terascale computers
• Create a Distributed Science Software Environment
  to enable management, dissemination, analysis
  of large data sets from simulation-intensive
  experimental/observational-intensive science

4
Centers for Enabling Technology

• Comprehensive, integrated, scalable, robust
  high performance computing software
  infrastructure to enable effective use of
  leadership class computing resources
• Comprehensive, integrated, scalable, and robust
  distributed computing technologies for
  experimental facilities and science
• Targeted efforts to integrate advanced applied
  mathematics and computer science technologies
  into selected applications projects

5
The Center for Enabling Distributed Science

• The Center for Enabling Distributed Science
  (CEDS) will address critical DOE requirements
  relating to the distribution, orchestration of
  computational analysis, and serving of the
  petabyte-scale data to be produced at
  experimental facilities and supercomputer centers
  over the next five years

6
Specific Goals

• Distribution
• Reliable, secure, and rapid movement of large
  quantities of data, subject to deadlines and
  resource constraints.
• Orchestration
• The orchestration of large-scale distributed
  computations to analyze this data reliably and
  correctly
• Serving
• The creation, provisioning, and operation of
  high-performance science services

7
Distribution

• Tools for secure, dependable, high-performance
  distributed data placement,
• enabling the reliable and scheduled replication
  of large quantities of data over wide area
  networks and
• addressing issues of firewall traversal,
  security, automated configuration, storage space
  management, and adaptation
• Work will include allocation of resources related
  to data placement
• end-to-end (network, buffer, disk bandwidth)
  capacity and storage space reservation and
  management

8
Orchestration

• Tools for facilitating the distributed analysis
  of large quantities of data, via the
• harnessing of distributed computing, storage, and
  network resources and the
• orchestrating of large-scale data analysis
  computations on those resources

9
Serving

• Tools for building the science services that are
  increasingly often being used to serve both raw
  data and data analyses to large user communities
• These services must enable
• high-performance access to, and processing of,
  large quantities of data
• dynamic deployment and provisioning in response
  to changing user demands
• service discovery
• monitoring for the purpose of problem diagnosis
  and service tuning

10
In Partnership With

• Open Science Grid!
• High energy and nuclear physics
• Other OSG application communities
• Earth System Grid
• Fusion, biology, nanosciences, astrophysics
• Other DOE application communities

11
Participants

• Argonne National Laboratory
• FermiLab
• Lawrence Berkeley National Laboratory
• USC Information Sciences Institute
• University of Wisconsin Madison

12
How You Can Help

• Tell us what you need most urgently
• The more details the better
• Put us in touch with additional DOE application
  communities