U'S' ATLAS Computing Facilities Overview

1
U.S. ATLAS Computing Facilities Overview

• US ATLAS Computing Advisory Panel Meeting
• October 25-27, 2000
• Bruce G. Gibbard, BNL

2
Facilities Presentations

• Overview B. Gibbard
• Computing requirements
• Network considerations
• Overall plan rationale
• Details of Tier 1 R. Baker
• Tier 1 configuration
• Cost schedule analysis
• Details of Tier 2s and Grid R. Gardner
• Tier 2 configuration
• Grid plan
• Cost schedule analysis

3
US ATLAS Computing Facilities

• Facilities procured, installed and operated
• to meet U.S. MOU obligations to ATLAS
• Direct IT capacity (Monte Carlo, for example)
• Support for detector construction, testing,
  calib.
• Support for software development and testing

• to enable effective participation by US
  physicists in the ATLAS physics program!
• Direct access to and analysis of physics data
  sets
• Simulation, re-reconstruction, and reorganization
  of data as required to complete such analysis

4
Setting the Scale

• Significant Uncertainty in Defining Scale of
  ATLAS Computing Needs
• Accelerator and detector far from completed
• Five years of algorithm software development
• Five years of computer technology evolution
• For US ATLAS
• Start from ATLAS Estimate of Requirements Model
  for Contributions
• Adjust for US ATLAS perspective (experience,
  priorities and facilities model)

5
US ATLAS Perspective

• US ATLAS facilities must be adequate to meet all
  reasonable U.S. ATLAS computing needs
  (Specifically, the U.S. role in ATLAS should not
  be constrained by a computing shortfall it
  should be enhanced by computing strength)
• There needs to be significant capacity beyond
  that formally committed to International ATLAS
  which can be allocated at the discretion of U.S.
  ATLAS among its analysis efforts

6
ATLAS Estimate (1)

• New Estimate Made As Part of Hoffmann LHC
  Computing Review
• Current draft ATLAS Computing Resources
  Requirements, V1.5 by Alois Putzer
• Assumptions for LHC / ATLAS detector performance

7
ATLAS Estimate (2)

• Assumptions for ATLAS data

8
Architecture

• Hierarchy of GRID Connected Distributed Computing
  Resources
• Primary ATLAS Computing Centre at CERN
• Tier 0 Tier 1
• Remote Tier 1 Computing Centers

• Remote Tier 2 Computing Centers
• Institutional Computing Facilities
• Individual Desk Top Systems

9
Functional definitions

• Tier 0 Center at CERN
• Storage of primary raw and ESD data
• Reconstruction of raw data
• Re-procession of raw data
• Tier 1 Centers (At CERN in some major
  countries)
• Simulation and reconstruction of simulated data
• Selection and redefinition of AOD
• User Analysis
• Tier 2 Centers
• Less clearly defined, reduced scale (
  functionality?) Tier 1

10
Required Tier 1 Capacities

• Individual Tier 1 Center Capacities in 2006
• CPU 209 kSPi95
• Disk 365 TBytes
• Tape 1800 Tbytes
• Expect 6 Such Remote Centers
• USA, FR, UK, IT, 2 to be determined
• Capacity Ramp-up Profile
• (Perhaps too much too early, as Rich Baker will
  discuss)

11
US ATLAS Facilities

• Requirements Related Considerations
• Analysis is the dominant Tier 1 activity
• Experience shows analysis will be compute
  capacity limited
• US scale is larger than other Tier 1 countries
  by authors, by institutions, by
  core fraction (x 1.7, x 3.1, x 1.8) and so will
  require more analysis than a single canonical
  Tier 1
• US effort within itself is more distributed while
  being more isolated from CERN than other current
  Tier 1 countries
• US Tier 1 must be augmented by additional
  capacity (particularly for analysis)
• Appropriate US facilities level is 2 x
  canonical Tier 1

12
US ATLAS Facilities Plan

• US ATLAS will have a Tier 1 Center, as defined by
  ATLAS, at Brookhaven
• The Tier 1 will be augment by 5 Tier 2 Centers of
  comparable aggregate capacity
• This model will
• exploit high performance US regional networks
• leverage existing resource at Tier 2 sites
• establish architectural support for the inclusion
  of institutional resources at other (non-Tier 2)
  sites
• focus on analysis both increasing capacity and
  encouraging local autonomy and creativity within
  the analysis effort

13
Schematic of Model
14
Facilities WBS
15
Tier 1 (WBS 2.3.1)

• US ATLAS Tier 1 is now operational with
  significant capacities at BNL in coordinating
  with the RHIC Computing Facility (RCF)
• There is broad commonality in requirements
  between ATLAS and RHIC and long term synergy with
  RCF is expected
• Personnel and cost projections for US ATLAS
  facilities are heavily base on recent experience
  at RCF see Rich Bakers talk

16
Tier 1 (continued)

• Full Tier 1 Functionality Includes ...
• Dedicated High Bandwidth Connectivity to CERN
• Primary Site for Storage/Serving
• Cache/Replicate CERN other data needed by US
  ATLAS
• Computation
• Primary Site for any US Re-reconstruction
  (perhaps only site)
• Major Site for Simulation Analysis
• Regional support plus catchall for those without
  a region
• Repository of Technical Expertise and Support
• Hardware, OSs, utilities, other standard
  elements of U.S. ATLAS
• Network, AFS, GRID, other infrastructure
  elements of WAN model

17
Tier 2 (WBS 2.3.2.7 - 2.3.2.11 )

• The standard Tier 2 configuration will focus on
  the CPU and cache disk required for analysis
• Some Tier 2s will be custom configured to
  leverage particularly strong institutional
  resources of value to ATLAS (the current
  assumption is that there will be 2 HSM capable
  sites)
• Initial Tier 2 selections (2 sites) will be base
  on their ability to contribute rapidly and
  effectively to development and test of this Grid
  computing architecture

18
Tier 2 (continued)

• Developmental Phase
• 1st Tier 2 will be selected and receive initial
  funding this year, FY 01
• 2nd Tier 2 will receive initial funding in FY 02
• These 2 developmental Tier 2 sites will
  participate in 2003 ATLAS MDC 2
• Deployment Phase
• Remaining 3 Tier 2s will be selected on the
  basis of a single proposal process in 2003 and
  will receive initial funding in 2004
• All Tier 2s will be fully operational for 2006
• See Rob Gardners talk

19
Network

• Tier 1 Connectivity to CERN and to Tier 2s is
  Critical to Facilities Model
• Must have adequate bandwidth
• Must eventually be guaranteed and allocable
  bandwidth (dedicated and differentiate) ?
• Should grow with need OC12 to CERN should be
  practical by 2005
• While the network is a integral part of the US
  ATLAS plan, its funding is not part of the US
  ATLAS budget

20
WAN Configurations
21
US ATLAS Capacities
22
Risks and Contingency

• Risk Factors
• Requirements may change
• Price/performance projections may be too
  optimistic
• Tier 2 funding remains less than certain
• Grid projects are complex and maybe be less
  successful than hoped

• Have Develop Somewhat Conservative but Realistic
  Plans and Now Expect to Build Facilities to Cost
• Contingency takes the form of reduction in scale
  (design is highly scaleable)
• Option to trade one type of capacity for another
  is retained until very late (80 of procured
  capacity occurs in 06)