Upload
preston-ramos
View
37
Download
0
Tags:
Embed Size (px)
DESCRIPTION
U.S. ATLAS Computing. Overview Status of ATLAS computing U.S. ATLAS Project Management Organization Status of efforts Core software Subsystems Facilities Schedule FY 00 funding request Summary. Scale of Computing Effort. - PowerPoint PPT Presentation
Citation preview
DOE/U.S. ATLAS Computing Sept. 9, 1999
U.S. ATLAS Computing
Overview Status of ATLAS computing U.S. ATLAS
Project Management Organization Status of efforts
Core software Subsystems Facilities
Schedule FY 00 funding request
Summary
DOE/U.S. ATLAS Computing Sept. 9, 1999
Scale of Computing Effort
Rough scaling of factors of 5 to 1E+3 in relevant parameters from Tevatron Experiments Manpower x5 CPU/event x1E+3 (event complexity) Data volume x10 to x1E+2 (channel count) Distribution of data x10
U.S. effort comparable to scale of Tevatron experiment.
Effort $15M/year
DOE/U.S. ATLAS Computing Sept. 9, 1999
Scales from experience
[email protected] : ATLAS Control : 01sep99 : ht tp://arc.nersc.gov/control/
HENP Computing Challenges
Ex perime nt Da ta ComputeE895 (AGS) 10 TB/yr 600 SPECint95BaBar (SLAC) 400 TB/yr 5,000 SPECint95STAR (RHIC) 266 TB/yr 10,100 SPECint95PHENIX (RHIC) 700 TB/yr 8,500 SPECint95D0 Run II (FNAL) 280 TB/yr 4,075 SPECint95CDF Run II (FNAL) 464 TB/yr 3,650 SPECint95ATLAS (LHC) 1100 TB/yr 2,000,000 SPECint95
Experiment CountriesInstitutes Collaborators Time FrameE895 (AGS) 3 12 49 2000BaBar (SLAC) 9 85 600 2010STAR (RHIC) 7 34 400 2010PHENIX (RHIC) 10 41 400 2010D0 Run II (FNAL) 11 77 500 2005CDF Run II (FNAL) 8 41 490 2005ATLAS (LHC) 34 144 1700 2015
DOE/U.S. ATLAS Computing Sept. 9, 1999
Goals for the next year
Project organization Management Identify areas of responsibility
Integration of efforts into ATLAS Inception/development of software U.S. support facilities
Planning/development of infrastructure
Prepare for reviews
DOE/U.S. ATLAS Computing Sept. 9, 1999
International ATLAS
New Computing Coordinator Norman McCubbin (RAL)
Available full time November Approval vote - ATLAS CB June 10th Responsibility: Core software
New Physics Coordinator Fabiola Gianotti (CERN)
Approval vote - ATLAS CB June 10th
Detector specific sim/reconstruction Organized within subsystems
DOE/U.S. ATLAS Computing Sept. 9, 1999
Architecture Taskforce
Software partitioned into work packages
Katsuya Amako, KEK Laurent Chevalier, CEA Andrea Dell’Acqua, CERN Fabiola Gianotti, CERN Steve Haywood, RAL (Chair) Jurgen Knobloch, CERN Norman McCubbin, RAL David Quarrie, LBL R.D. Schaffer, LAL Marjorie Shapiro, LBNL Valerio Vercesi, Pavia
DOE/U.S. ATLAS Computing Sept. 9, 1999
Architecture T.F. Status
Three meetings so far Directions:
Language: C++ (allow for migration to other e.g. JAVA) Examine GAUDI (LHCb) architecture Adoption of “use cases”
Goals for October Outline of architecture design Appointment of Chief Architect Commission work on prototyping of parts of design Create use-cases, requirement document Define packages and relations (package diagram)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Stages in Software Management
Atlas Software Week 1999.09.01 K.Amako 20
Software Development Process: USDP
Workflows vs. Development Phase - Iterative and incremental [USDP p.11]
ManagementEnvironment
Business Modeling
Implementation
Test
Analysis & Design
Preliminary Iteration(s)
I ter.#1
PhasesProcess Workflows
Iterations
Supporting Workflows
I ter.#2
I ter.#n
I ter.#n+1
I ter.#n+2
I ter.#m
Iter.#m+1
Deployment
Configuration Mgmt
Requirements
Elaboration TransitionInception Construction
DOE/U.S. ATLAS Computing Sept. 9, 1999
Quality Control
Recommend software performance specifications, review process
Makoto Asai, Hiroshima Dario Barberis, Genoa Martine Bosman, Barcelona Bob Jones, CERN Jean-Francois LaPorte, CEA Helge Meinhard, CERN Maya Stavrianakou, CERN
DOE/U.S. ATLAS Computing Sept. 9, 1999
Action on other Groups
National Board Supported platforms Regional centers
Training Network of national contacts for training C++, OO programming GEANT 4 ATLAS Specific
DOE/U.S. ATLAS Computing Sept. 9, 1999
ATLAS/CERN Schedule ‘00
Sept. ‘99 Start of Cashmore/Hoffman review
Oct. ‘99 Report of architecture T.F. Commissioning of prototyping code
Jan ‘00 Start preparations for bilateral agreements
Fall ‘00 Report of Cashmore/Hoffman review MOU preparations
DOE/U.S. ATLAS Computing Sept. 9, 1999
U.S. Participation Frank Paige - Co- convenor of SUSY working group David Malon - Co-leader of database group Craig Tull - Architecture Group Ian Hinchliffe - Leader of Event Generator group David Quarrie, Marjorie Shapiro - Architecture Task Force John Parsons - Co-convenor of Top working group Misha Leltchouk - L Ar simulation coordinator Michael Shupe - Convenor of Background working group Fred Luehring - TRT software coordinator Steve Goldfarb - Muon Database Coordinator Tom LeCompte - Tilecal Database Coordinator Krzys Sliwa - Chair of ATLAS World-wide computing group Frank Merritt - Training contact, Tilecal Reconstruction coord. Bruce Gibbard - Regional center contact John Huth- National Board contact
DOE/U.S. ATLAS Computing Sept. 9, 1999
U.S. ATLAS Computing
NSF, DOE: LHC computing activities are now “projectized”
Implications for U.S. ATLAS: Direct reporting lines through Project
Manager (Bill Willis) and BNL Directorate (Tom Kirk)
Appointment of Associate Project Manager for Computing and Physics (John Huth)
Implications for Agencies: Must clarify reporting lines, operations
DOE/U.S. ATLAS Computing Sept. 9, 1999
Proposed Management
William WillisProject Manager
William WillisProject Manager
John HuthAssociate Project
Manager,Computing and
Physics WBS 2
James ShankDeputy
External AdvisoryGroup
Ian HinchliffeManager, Physics
WBS 2.1
TBN1Manager,SoftwareWBS 2.2
Bruce GibbardManager,FacilitiesWBS 2.3
Craig TullControl/
FrameworkWBS 2.2.1.1
David MalonDabase Systems
WBS 2.2.1.2
Srini RajagopalanEvent ModelWBS2.2.1.3
F. Merritt/J.Shank
Detector SpecificWBS2.2.2
F. MerrittTraining
WBS 2.2.4
Laurent VacavantInner DetectorWBS 2.2.2.1
Keith BakerTRT
WBS 2.2.2.2
Srini RajagopalanEM Cal.
WBS 2.2.2.3
Tom LeCompteTilecal
WBS 2.2.2.4
Bing ZhouMuons
WBS 2.2.2.5
Andy LankfordTrigger/DAQWBS 2.2.2.6
SubsystemsCore Software
S. EfstathiadisSoftware Support
WBS 2.3.1.2
AshkanaziHardware/systems
WBS 2.3.1.1
TBNRemote Sites
WBS 2.3.2
TBN(Price)NetworkingWBS 2.3.3
K. SliwaMONARCWBS 2.2.4
Facilities
TBNCollaborative
ToolsWBS 2.2.3
DOE/U.S. ATLAS Computing Sept. 9, 1999
Management Structure
Reflects flow of deliverables to, from ATLAS
Appointments (2 year renewable terms) Physics: Ian Hinchliffe (LBNL) Facilities: Bruce Gibbard (BNL) + deputy
Issues Software manager
Availability within U.S. ATLAS - hire? Flatter structure for the time being? Bring on soon? Most desirable!
DOE/U.S. ATLAS Computing Sept. 9, 1999
Management Plan
Associate Project Manager Member of E.C. Develop and execute project plan Establish and maintain project organization+Tracking Develop annual budget requests Liason to ATLAS Computing Management Appoint L2 managers Review and approve MOU’s to CERN and Institutes Exercise change control authority Establish advisory committees where appropriate Provide reports and organize reviews
DOE/U.S. ATLAS Computing Sept. 9, 1999
Implications for APM
APM is a time consuming job. Actions for John Huth:
Relief from MDT electronics coordinator (Jay Chapman, U. Michigan)
Relief from U.S. ATLAS Inst. Bd. Chair (Jim Siegrist, LBNL)
Teaching relief - spring terms ‘00 and ‘01 Granted by Harvard University
DOE/U.S. ATLAS Computing Sept. 9, 1999
Level 2 Managers
Appointed by APM, concurrance of Exec. Comm. Members of E.C. (+ APM, + deputy) Two year renewable terms Generic responsibilities
Develop definition of milestones and deliverables Define, with APM, organizational substructure of level 2 Develop, with APM, annual budget proposals Identify resource imbalances within subprojects and
recommend adjustments Deliver scope of subproject on time within budget Maintain cost and schedule Provide reports to APM, PM Liason with counterparts at CERN
DOE/U.S. ATLAS Computing Sept. 9, 1999
Specific Responsibilities
Physics Manager Generators, physics objects, benchmark
studies, mock data challenge
Software Core Detector specific sim/recon Training
Facilities Tier 1,2, networking, support
DOE/U.S. ATLAS Computing Sept. 9, 1999
Project Engineer
Same roles as project engineer’s for construction project Tracking Reviews, oversight Reporting Technical input
DOE/U.S. ATLAS Computing Sept. 9, 1999
Proposed Names
Physics: Ian Hinchliffe (LBNL) Contacted, accepted Policy question: physicists on project
Software: search underway Facilities: Bruce Gibbard (+deputy) Deputy: Jim Shank
DOE/U.S. ATLAS Computing Sept. 9, 1999
Facilities Manager
Bruce Gibbard (BNL) Proposal to add U.S. ATLAS Deputy U.S. ATLAS and RHIC Deputies General agreement with Bruce, Tom Kirk
Begin to fill in other areas Networking Tier 1 Remote sites Support
DOE/U.S. ATLAS Computing Sept. 9, 1999
Detector Contacts
L Ar - Srini Rajagopalan (BNL) Tilecal - Frank Merritt (U.Chicago) ID- Laurent Vacavant (LBNL) Muon - Bing Zhou (U. Michigan) TRT - Keith Baker (Hampton) Trigger/DAQ -Andy Lankford (UCI)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Planning Activities
Writing/preparation assignments for review - next week L2 managers where appropriate (ie. Facilities,
physics) Core Sim/recon Training Management (PMP for computing) MRE/IT “team”
Review in October
DOE/U.S. ATLAS Computing Sept. 9, 1999
WBS Structure
Should be flexible while project definition is underway (level 3+beyond)
Level 2’s should be fixed now Commensurate with management
structure Adopt lead number “2”
DOE/U.S. ATLAS Computing Sept. 9, 1999
High Levels of WBS
Draft WBS 2.1 Physics
Generators, benchmarks, mock data challenges, physics objects
2.2 Software 2.2.1 Core
– Control/Framework,database, event model, analysis tools 2.2.2 Detector specific simulation and recon. 2.2.3 Collaborative tools 2.2.3 Training
2.3 Facilities Regional center, remote sites, networking, support
DOE/U.S. ATLAS Computing Sept. 9, 1999
Near Term Activities/Issues
U.S. ATLAS Web-site Weekly video conferences Support role of BNL Gathering FY 00 requests Advisory group appointment Writing assignments for proposal NSF MRE/IT proposal - Tier 2 centers Discussions of deliverables with ATLAS Interactions with agencies
JOG, Computing review
DOE/U.S. ATLAS Computing Sept. 9, 1999
Software
Core Software Control/Framework (Tull) Database, Tilecal Pilot Project (Malon) Event Model (Rajagopalan)
Detector-specific sim/reconstruction Representatives from subsystems chosen
Training (Merritt) Establishment of OO courses (BNL, U.
Chicago)
DOE/U.S. ATLAS Computing Sept. 9, 1999
General Requirements
Software must last over lifetime of experiment, yet track language changes Well defined interface layers
Maintainability, engineering critical Number of users, use of software
professionals
Adaptability to distributed environments
Learn from experiments working on OO (BaBar, D0, CDF, STAR)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Database
David Malon (ANL) Tilecal pilot project
Tilecal testbeam data in object database Testbed for ATLAS technologies and strategies Early feedback to developers Generalized to other subsystems
Database core software Transient and persistent object mapping Definition of database/control interface Specifications Examine alternatives to Objectivity
DOE/U.S. ATLAS Computing Sept. 9, 1999
Tilecal Model
Argonne National Laboratory
Tilecal test beam analysis (M. Nessi)
Data base
DA QGeant4
Geant3
Interface via tile logical description
DataDump Calibration
analysis
Optimalfilteringanalysis
Histogramming
Physics analysis
JAVABrowser
...
DOE/U.S. ATLAS Computing Sept. 9, 1999
Database Schedule
DOE/U.S. ATLAS Computing Sept. 9, 1999
Database Milestones
Jan ‘00 Survey of mapping strategies
Feb. ‘00 Infrastructure for developers deployed
April ‘00 Validation of strategies in testbed
July ‘00 Database management infrastructure defined
Oct ‘00 Infrastructure for distributed access deployed
Jan ‘01 Scalability test
Oct ‘01 Beta release
DOE/U.S. ATLAS Computing Sept. 9, 1999
Control/Framework
Craig Tull (LBNL) Working on user requirements document
(w/ Hinchliffe, Shapiro, Vacavent) Market survey of framework systems
Object component model AC++
Compatibility with ATLAS architecture Resource loaded work plan exists Work with A.T.F. for design requirements Already have tested prototype designs
DOE/U.S. ATLAS Computing Sept. 9, 1999
Framework Milestones
Task NameDraft reqs. complete
Compl. initial framework survey
Req. document complete
Requirements review
Freeze CORE language
Alpha release design review
Alpha release
Freeze CORE architecture
Freeze database interface
Freeze platforms for MDC
Beta release design review
Beta release
Freeze distributed architecture
Vers. 1.0 design review
Vers. 1.0 release
MDC start
Freeze platforms for production
MDC complete
Production release review
Production release
Data taking starts
8/27
2/1
2/1
3/1
3/29
6/29
9/28
3/29
3/29
6/4
7/2
10/1
4/1
7/1
9/30
6/2
7/4
1/2
4/5
7/2
7/1
2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 19 2000 2001 2002 2003 2004 2005 20
DOE/U.S. ATLAS Computing Sept. 9, 1999
Framework Schedule
Task NameProject Management
Developer Support
Define requirements
Survey existing frameworks/arch
Survey technologies
Draft reqs. complete
Evaluate dictionary language
Evaluate core language
Domain decomposition
Development tools
Compl. initial framework survey
Req. document complete
Requirements review
Freeze CORE language
Code generation tools
Flow control
Analysis tools interface
Document Alpha release
Alpha release design review
Test Alpha release
Alpha release
Database interface design/proto
Freeze CORE architecture
Freeze database interface
Tull[10%]
Milford[50%]
Tull[15%],Hinchliffe[10%],Shapiro[10%],Vacavant[10%]
Calafiura[15%],Leggett[15%],Vacavant[15%]
Tull[10%],Calafiura[10%],Milford[10%]
8/27
Tull[15%],Calafiura[15%]
Calafiura[15%],Leggett[15%]
Tull[50%]
Calafiura[33%],Vacavant[33%],Leggett[33%]
2/1
2/1
3/1
3/29
Calafiura[20%],Milford[40%],TBN[20%]
Tull[50%],TBN[50%]
Leggett[33%],Vacavant[33%]
All
6/29
All
9/28
TBN[75%]
3/29
3/29
4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 12000 2001 2002 2003 200
DOE/U.S. ATLAS Computing Sept. 9, 1999
One Framework Model
Software Bus (eg. CORBA)
Component C++ Classes/Objects
Component Class Adatpers
Scripting Interface (eg. Tcl, …)
Command Marshalling (eg. SWIG, ...)
GUI Interface (eg. Tk, …)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Event Model
Type of objects that are stored Client’s view of the event Physical organization of the event Client’s interface to the event Mechanism to navigate between objects Transient to Persistent Object mapping
DOE/U.S. ATLAS Computing Sept. 9, 1999
Event Model Milestones
Nov. 99 Review of models in other exp’s Requirements documents
Dec. 99 Resource loaded schedule
Mar. 00 Baseline design
June 00 Alpha release
Aug. 00 Review experience
FY 01 Beta release
DOE/U.S. ATLAS Computing Sept. 9, 1999
Framework Options
[email protected] : ATLAS Control : 01sep99 : ht tp://arc.nersc.gov/control/
Framework Design Classifications
• Finite State Machine - AC++
• Action on Demand - CARF
• Stream/Record/Frame - CLEO
• Simulated Data Flow - Gaudi
• Mobile Agents - JAS
• Object Network - ONCM
• C++ Interpreter - ROOT
• Software Bus - StAF
DOE/U.S. ATLAS Computing Sept. 9, 1999
Some Detector Activities
TRT/ID Put full TRT
simulation into GEANT4
L-Ar Coil, cryos in
GEANT4 (Nevis) Accordian structure
in GEANT4 (BNL)
Tilecal Pilot project
DOE/U.S. ATLAS Computing Sept. 9, 1999
Some Detector Activities
Muon Study of noise in Higgs-> 4 muon Combined performance of ID+muon system
(A reconstruction) CSC into simulation
Trigger/DAQ Comparison of switching architectures
Background studies Optimization of shielding (100 MeV muon
background)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Training
New paradigm of OO programming Training courses (F. Merritt)
Course offered at BNL (near future) Course offered at Chicago
Successful programs seen at other experiments (CDF, D0, BaBar)
Ongoing need for training throughout course of experiment Documentation ATLAS-specific
DOE/U.S. ATLAS Computing Sept. 9, 1999
Software Development
Asymptotic level - est. 10 software professionals
Peak load (circa 2003) est. 16 S.P.’s Extrapolations based on existing
experiments and proposed areas of responsibility, fractional of U.S. participation
Choice of technology can influence actual needs strongly (e.g. BaBar, STAR in database)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Planned Training
All by Object Mentor (BaBar, others) Organized by Frank Merritt Courses approximately 1 week long Aug. 9 - BNL - OO Design - 13 people Sept. 20 - U.C. - OO Design - 15 people Oct. 18 - ANL or BNL - Advanced OO -
10 people Nov. 8 - FNAL - GEANT 4 - 14 people
DOE/U.S. ATLAS Computing Sept. 9, 1999
Facilities
BNL ramping up support facility Taps into RHIC Computing Facility
Major issue of Tier 1/2 facilities Scale of “Tier 2’s”
Size for support staff, infrastructure Computing model for U.S. (e.g. grids) Being addressed in NSF MRE/IT proposal
In the process of developing policy on usage, support of platforms at institutions
DOE/U.S. ATLAS Computing Sept. 9, 1999
Facilities
Tier 1 (Regional Center) BNL Leverages RCF
ATLAS specific needs, however. Primary support function for U.S.
Code release, support Major processing, event store
Personnel scale estimate: Roughly linear ramp from 2 FTE’s (now) to 22 or
more (depending on computing model)
DOE/U.S. ATLAS Computing Sept. 9, 1999
MONARC
Models of Networked Architecture at Regional Centers (ATLAS+CMS) Alexander Nazarenko, Tufts hire Tasks:
Validate simulation models Perform first simulations of LHC architectures After Dec. ‘99, focus on planning for regional
centers Model validation - end of September
Understanding of U.S. computing facilities
DOE/U.S. ATLAS Computing Sept. 9, 1999
NSF MRE/IT Proposal
Tier 2 centers Approx. 5 total 256 node systems 100 TB tape system Low maintenance
Linked by computing grid
Computing professionals Dual role -
user/developers
DOE/U.S. ATLAS Computing Sept. 9, 1999
Review Process
Send proposals to Advisory Group (Aug. 4th) Charge:
Identify areas of overlap/commonality Suggest simplifications/savings Coherency with ATLAS effort Prioritize
Meet with Agencies to establish scale for FY 00, initial request (Now)
Confer with Advisory group on feedback Prepare, review documentation for Dec.
review (mid-late Sept.)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Priorities
Critical personnel People who would otherwise be lost, fulfilling a
critical role
Core software effort Prerequisite to inclusion of sim/recon software Yet, cannot commit to major ramp (no MOU’s)
Support of U.S. efforts (facilities) Critical studies Transition to OO
DOE/U.S. ATLAS Computing Sept. 9, 1999
Priorities
Coherency in development of plan Matching of facilities scope to usage
E.g. database effort, simulations Contiguous/overlapping areas
E.g. event model, database, control/framework
DOE/U.S. ATLAS Computing Sept. 9, 1999
FY 00 Needs
Starting Oct. 1st - will need more than simple continuation of present level Support functions at BNL Deputy Facilities Manager Core support - database
Estimate roughly 4 FTE increment Remaining needs made part of review
process for Dec. Still give estimates for needs beyond
December -> now
DOE/U.S. ATLAS Computing Sept. 9, 1999
Facilities
Support for U.S. users a necessary precondition for effective U.S. participation (like training)
Use of RCF leverages existing facilities Requesting 3 FTE’s Oct. 1
(deputy+support) $50K of equipment Oct. 1 (500
SpecInt95) $500K (CPU, disk, mass storage) - after
review
DOE/U.S. ATLAS Computing Sept. 9, 1999
Training
Necessary precondition to effective U.S. participation.
Must be done now (trained group of physicists)
Substantial pay-back (experience in industry)
NSF request of $75K to subsidize courses ($1k/student + setup)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Leveraging Funds
Groups that have applied for internal funds BNL: Already support, 2 FTE’s for FY 00
(core software) LBNL: Request for support on
control/framework U.Chicago: Advance support for training,
request for software professional to start
DOE/U.S. ATLAS Computing Sept. 9, 1999
FTE Request
Institute ‘99 Oct. 1 Feb. 1 Sum
ANL 1 0.5 1 2.5
BNL 0 2.7 2 4.7
LBNL 1.4 0.8 1 3.2
U.C. 0 0 1 1
Columbia 0.5 0.5 0 1
Tufts 1 0 0 1
Total 3.9 4.5 5 13.4
DOE/U.S. ATLAS Computing Sept. 9, 1999
Schedule
July Propose management structure to E.C., PM Collaboration meeting Tier 1/2 scoping Plans for FY 00 reviewed MRE “White paper”
August Present FY 00 plans to agencies Outline and writing assignments for proposal
(Dec.)
DOE/U.S. ATLAS Computing Sept. 9, 1999
Schedule
September First drafts of proposal
Management PMP Software: Core and recon/sim Facilities Training, collaborative tools
October Revise proposal, review
November Meeting to prepare for Dec. review
DOE/U.S. ATLAS Computing Sept. 9, 1999
Schedule
December (January?) Agency review
January Revise funding plan for FY 00 Begin work on bilateral agreements Ongoing - and beyond January Prototyping code Progress toward baselining Filling in management slots Bilateral agreements
DOE/U.S. ATLAS Computing Sept. 9, 1999
Summary
Project organization Management Identify areas of responsibility
Integration of efforts into ATLAS Inception/development of software U.S. support facilities
Planning/development of infrastructure
Prepare for reviews
DOE/U.S. ATLAS Computing Sept. 9, 1999
Summary
Major points: Oct. 1st FY 00 needs are more than
ongoing (4 FTE) Hardware to augment RCF Training physicists in OO design Continue to fill in management structure