SNOLAB DataGrid

1
SNOLAB DataGrid

• This presentation will probably involve audience
  discussion, which will create action items. Use
  PowerPoint to keep track of these action items
  during your presentation
• In Slide Show, click on the right mouse button
• Select Meeting Minder
• Select the Action Items tab
• Type in action items as they come up
• Click OK to dismiss this box
• This will automatically create an Action Item
  slide at the end of your presentation with your
  points entered.

• Alain Bellerive
• Canada Research Chair

2
OUTLINE

• Introduction
• - The Scope of Particles Astrophysics
• - Sudbury Neutrino Observatory (SNO and SNOLAB)
• - Data Flow and Partnership with ORION
• - Identify Global Networking Needs
• SNOLAB DataGrid
• - Activities at Carleton University and in
  Sudbury
• - Data Distribution Support to International
  Collaborators
• Summary

3
Scope Particle Physics

• Study fundamental forces of nature and its
  elementary constituents.
• The neutrino is the second most abundant particle
  in the Universe.
• Use the sun to investigate the properties of weak
  interacting particles called neutrinos !
• The earth receives around 100 billions neutrinos
  per second and cm2.

4
Solar Neutrino Problems
Light Element Fusion Reactions
p p ?2H e ?e
99.75
p e- p ? 2H ?e
0.25
10-5
3He p ?4He e ?e
7Be e- ?7Li ?e
15
?e(Measured)? 1/3 ?e(Model)
8B ? 8Be e ?e
0.02
Neutrino Production in the Sun
?e oscillate to ??? as discovered by SNO BIG
SCIENCE
5
Data Flow for SNO

• Raw Data
• Rate for neutrino runs is 15 Hz 2 Kb/event
• Rate for calibration runs is 50-200 Hz lt200
  Kb/event
• Label the data by runs (i.e. time) and running
  conditions
• Each year SNO records around 1Tb of raw data.
• Event Summary
• Reduced data set from pattern recognition
• Event summary files used for data analysis
• Correspond to another 200 Gb per year
• Monte Carlo (MC) Simulation
• Large volume of MC events to be distributed
  world-wide
• Scale by a factor of 20 or 400 the amount of
  simulated data compared to the raw data collected
• MC data required correspond to 1-3 Tb per year

6
Early Partnership with ORION

• In 2001, the network limitation was the slow
  connectivity via the 7 Mb/sec ATM connection
  between Toronto ? Sudbury
• Major improvement in May 2003 with connection to
  ORION THANKS!
• Led to quick turn-around for diagnostics, data
  analysis, and publication

7
CFI International Venture SNOLAB
New cavern at the 6800 ft level !!!
8
SNOLAB Requirements

• The present data flow from SNO will only be a
  part of the data expected from the new SNOLAB
• For the design of experiments for this facility,
  there will be substantial Monte Carlo simulations
  required and for the processing of data a further
  requirement will exist beyond the present SNO
  needs
• Need to support at least three major experiments
  to study
• Low energy solar neutrinos
• Double Beta decay
• Search for Dark Matter
• In the future, SNO and SNOLAB, will benefit
  substantially from a high-speed national network
  CANet4 ORION dedicated to science

 
9
Connectivity Raw Data Computing

• Fast Ethernet switches (Gb/s) and optical fibres
  connection from underground to the surface
  building Thanks for matching funds from ORION
• A total of 10-100 Tb of data needs to be
  accessible by all the member institutions of
  SNOLAB in Ontario
• Not a lot of data (compared to others experiments
  in HEP)
• Need to move the data out of Sudbury
• Particle physics does not share the need of
  parallel computing like in other discipline HEP
  requires a stable and users oriented platform
• Focus on throughput computing
• Some calculations need to performed on-site, but
  the majority will be done remotely
• Users and computing resources are scattered
  around the world

Connection underground ? surface (ok)
Data Flow (manageable)
Computing (distributed resources)
Network (challenge)
10
SNOLAB DataGrid
HPC
GRID
GridComputing (solution)
NETWORK

• Advances in distributed computing, high quality
  networks, and powerful and cost-effective
  commodity-based computing have given rise to this
  Grid computing
• Grid projects present fresh challenges for a
  number of reasons, such as the difficulty of
  coordinating the use of widely distributed
  resources owned and controlled by many
  organizations
• The Grid introduces the concept of the virtual
  organization as a group of both users and
  computing resources from a number of real
  organizations that are brought together to work
  on a particular project
• The objective of the SNOLAB DataGrid project is
  to assist the next generation of scientific
  exploration, which requires intensive computation
  and analysis of shared large-scale datasets
  across universities in Ontario e.g. Queens
  Carleton Laurentian Guelph.
• Need support for the development of
  infrastructure for distributing data across large
  area networks (i.e. people involvement)

11
Partnership and Challenges

• Particle physicists are concerned with the
  management of large data samples
• Benefits from the dedication of ORION to
  interconnect Carleton, Queens, Laurentian, and
  SNOLAB
• Development of point-to-point light-path
  connection in partnership with CANARIE
• Rely on the Globus Tool Kit developed for
  advanced distributed computing in High Energy
  Physics experiments
• SNOLAB main objective is the DATA aspect (not
  computational)
• Identify common research area with partners in
  Ontario
• Stimulate development of the framework needed for
  the data movements foresee in the next wave of
  network development
• Gain expertise for future Grid applications in
  particle physics (LHC) and other sectors
  (banking, stock market, medical imaging, etc)
• Call for large/secure/reliable data storage over
  a wide network of collaborators in Ontario and
  around the world

12
Data Distribution Architecture

• Progress towards research and deployment of a
  DataGrid at SNOLAB
• Alain Bellerive (Physics/Carleton)
• Aksel Hallin (Physics/Queens)
• Wade Hong (Physics/Carleton)
• Shikharesh Majumdar (Syst. Eng./Carleton)
• Clarence Virtue (Physics/Laurentian)

GridComputing (solution)
SNOLAB
Carleton
Queens
Laurentian
13
Summary
Leadership of Canada in Underground
Science Goal International leadership of SNOLAB
in Particle Astrophysics Opportunity Establish
SNOLAB indisputable scientific authority in the
fields of elementary particle physics and the
evolution of the Universe Benefits Investment in
academic training and forefront research  
DataGrid Role Distribute the data collected at
SNOLAB Research Discussions on fundamental
science is stimulated with the fast connectivity
between physicists and the need for data
management and analysis in collaborative effort
Infrastructure Development of the tools for
wide data distribution  
 
 

• Framework and Prove of concept at SNOLAB
• Gain expertise now in direct research
  involvement in data movement
• Production prototype on real data in preparation
  for LHC
• Workflow of Open Grid Service Architecture
  (OGSA) towards
• Web Services
•