Particle Physics Data Grid Scientific Discovery through Advanced Computing Collaboratory Pilot, Trillium

1
Particle Physics Data Grid Scientific Discovery
through Advanced Computing Collaboratory Pilot,
Trillium High Energy Physics Intergrid
Coordination and Joint Technical Boards

• LHC Review, January 14th
• R. Pordes, Fermilab,
• a PPDG Coordinator
• co-chair of the HICB-Joint Technical Board

http//www.ppdg.net/docs/presentations/PPDG_HICB_0
11403.ppt
2

• PPDG
• Introduction
• Management structure, coordination with the
  experiments
• Status and deliverables
• Prospects
• Trillium
• Introduction,
• Prospects
• HICB and HICB-JTB
• Process, Goals, Contacts with the Experiments
• Status and deliverables
• Prospects

• www.ppdg.net
• Q3-02 Qtrly Report
• Year 2 Plan 
• Analysis Tools meeting (12.02)
• Site Authentication and Authorization Report
  (12.02)
• Joint Collaboration Meeting
• Newsletter
• US-EU Demos
• www.hicb.org
• www.hicb.org/jtb
• www.hicb.org/GLUE

3
Reference urls

• www.ppdg.net
• Q3-02 Qtrly Report - http//www.ppdg.net/docs/ppdg
  _qtrly_jul-sep-02.pdf
• Year 2 Plan  - http//www.ppdg.net/docs/Project/PP
  DGYear2.doc
• Analysis Tools meeting (12.02) -
  http//www.ppdg.net/mtgs/19dec02-cs11/
• Site Authentication and Authorization Report
  (12.02) - http//www.ppdg.net/mtgs/19dec02-siteaa/
  
• Joint Collaboration Meeting - http//www.ivdgl.or
  g/events/view_agenda.php?id3
• Newsletter - http//www.hicb.org/TrilliumNewslette
  r/
• US-EU Demos http//www.ivdgl.org/demo/
• www.hicb.org
• www.hicb.org/jtb
• www.hicb.org/GLUE
• http//www.hicb.org/glue/glue-v0.1.3.doc

4
Introduction to PPDG

• DOE SciDAC Collaboratory Pilot. Currently in the
  middle of 2nd year of 3 year current program.
• PIs Miron Livny, Richard Mount, Harvey Newman.
  Executive Team Doug Olson, Miron, Ruth
• Annual budget 3M for effort - no funding for
  hardware. Funding split between MICS and HENP
  offices.

• Computer Science Groups
• Globus
• Condor
• SDSC
• LBNL SRM
• 1-3 FTEs each.

• Running Experiments
• Nuclear Physics Star, JLAB 1 FTE each
• HEP BaBar, D0 2-3 FTE each
• LHC Experiments ATLAS, CMS 2-3 FTE each
• LQCD connection JLAB

5
PPDG - Scope and Goals - the Green Grid

• Deploy End-to end-Experiment application Grids -
  work and integrate with the local environment.
• Integrate common grid technologies and standards
  to have working end-to-end systems - ensure
  useful work.

• Development integration Projects are peered
  collaborations of 1 CS and 1 Experiment group -
  with exceptions to prove the rule.
• Cross cut technology Common Services
  strategically defined to encourage
  cross-experiment and cs group discussion and
  adoption.
• Leverage wider efforts - think globally, act
  locally.
• Technical Challenges - Deployment is different
  than demonstration
• For the running experiments constrained to mesh
  with legacy systems Need to maintain operational
  current data handling and even Distributed Data
  Managements infrastructures throughout the
  lifetime of the project allow only tiny steps to
  be taken and even those take a long time.
• For LHC experiments and LQCD data handling models
  are still being developed in large
  collaborations. Parallel solutions and divergence
  easy.
• Sociological Challenges of course.

6
PPDG Work Areas

• Data Movement focus includes Reliability and
  Robustness up front.
• Experiment Production - simulation,reconstruction
  - Grids were an early primary deliverable -
  only at 10 because RELY on CS-1 through CS-9.
• Steering Committee extends the areas of work as
  we need to.
• Focus is moving down the list as the project
  proceeds.
• Clearly PPDG is not developing all these
  capabilities. We discuss and address application
  needs and technology extensions in all these
  areas.

7
Who participates

• Globus/ANL, Ian Foster , Mike Wilde , Jenny
  Schopf , Willliam Alcock, Von Welch, Stu Martin
• ATLAS, John Huth, Torre Wenaus, L. Price , D.
  Malon , A. Vaniachine, E. May , Rich Baker , Alex
  Undrus  , Dave Adams, Wengshen Deng, G.
  Gieraltowski , Dantong Yu
• BaBar, Richard Mount, Tim Adye , Robert Cowles ,
  Andrew Hanushevsky , Adil Hassan , Les Cottrell,
  Wilko Kroeger
• CMS, Lothar Bauerdick, Harvey Newman , Julian
  Bunn , Conrad Steenberg , Iosif Legrand ,
  Vladimir Litvin, James Branson , Ian Fisk, James
  Letts, Eric Aslakson, Edwin Soedarmadji, Saima
  Iqbal, Suresh Man Singh, Anzar Afaq, Greg Graham
• D0, Lee Leuking , Igor Terekhov , Andrew
  Baranovski, Gabriele Garzoglio, Siddharth Patil,
  Abhishek Rana, Hannu Koutaniemi, John Weigand
• HRM/LBNL, Arie Shoshani  , Alex Sim  , JunminGu,
  Alex Romosan
• SRB/UCSD, Reagan Moore , Bing Zhu
• JLAB, William Watson , Sandy Philpott, Andy
  Kowalski , Bryan Hess , Ying Chen, Walt Akers
• STAR , Jerome Lauret, Gabrielle Carcassi, Eric
  Hjort, Doug Olson
• Condor/U.Wisconsin, Miron Livny , Paul Barford ,
  Peter Couvares, Rajesh Rajamani, Alan DeSmet,
  Alain Roy,Todd Tannenbaum
• Globus/ISI, Carl Kesselman, Ann Chervenak
• SiteAAA(if not listed above), Dane Skow, Bob
  Cowles, Booker Bense, Tomasz Wlodek , Igor
  Mandrechenko, Steve Chan, Steve Lau , Craig Tull,
  Matt Crawford, David Bianco, Hao Wang

8
PPDG Management and Day to Day Operation

• PPDG funded effort is managed by Team Leads (or
  designees), in collaboration with the site or
  project leaders. Well integrated into their wider
  teams.
• Experiment Team Leads are Computing Managers.
  Computer Science Team Leads are CS group leaders.
• Project Business conducted through weekly
  Executive and Collaboration phone meetings
  review common service areas e.g. file replication
  and cover end-to-end status of experiment
  applications. 2 collaboration several focus
  meetings a year
• Information Flow through comprehensive quarterly
  reports, mail lists, web pages
• Steering Committee phone meetings monthly and
  face to face meetings 2 per year.
  Construct, review and endorse yearly plan. Take
  decisions on budget and strategy.
• Outreach and Liaison get much attention travel,
  coordination, communication and outreach to peer
  projects. Liaisons from iVDGL, GRiPhYN, LCG,
  International etc.

PIs
Steering CommitteePIs, Executive Team, Team
Leads, Liaisons

• Executive Team

9
Where PPDG Experiment Effort is focussed

• Experiment highest priority should be directly
  related to development and support for the
  integration and delivery of working grid
  applications in the experiment - e.g. Anzar Afaq
  in CMS is the coordinator of the Integration
  Production Testbed Gerry Giertowski in ATLAS
  evaluated and integrated European Data Grid
  software with production interface.
• Computer Science highest priority should be
  directly related to development and support for
  the identified short term requirements of the
  experiment applications. e.g.Globus changes to
  CAS in response to SiteAA requirements Condor
  extensions to DAGMAN to support CMS production
  testbed
• Development are most appropriate in areas where
  existing or emerging grid technologies are
  lacking, and the experiments have short to medium
  term needs. E.g. CMS and ATLAS developments for
  physics analysis SRB support for federations
• Outreach Work with projects and standards bodies
  e.g. Storage Management Interface definitions
  e.g.SRM Vx. work with all on proposing
  development of missing capabilities for
  production grids. e.g. with DOE SG on US CA/RA.

10
Status and Deliverables

• All experiments have deployed applications using
  Grid Data Movement technology for the benefit of
  their collaborations
• Data transfer GridFTP, BBCP SRM(DRM),
• Data movement management SAM, SRB, MAGDA,
  GDMP,

• Most experiments have prototype/development
  applications with Grid Job Scheduling and
  management Condor-G job scheduling to multiple
  Globus-Gatekeepers and Sites Resource discovery
  and monitoring through MDS DAGMAN for job flow.
• Cross project additional effort in Authentication
  and Authorization yielding working solutions
  where initially there were none.
• Many specific developments from experiment
  engineers to achieve working applications.

11
Year 2, Year 3,.. deliverables

• File Transfer, Replication and Data Management -
  in particular continued integration and
  deployment in the applications.
• Scheduling and management of physics data
  processing and analysis applications. This brings
  with it the need to address the details of
  intelligent specification, request, execution and
  presentation of work.
• Authorization - how application resource
  requests coming from the grid get mapped to the
  allocation and use of local facility resources,
  how the actual use is reported back to the
  authorizing entity.
• Production Capabilities Integration and
  Deployment, System Operation, Qualities for
  Production
• Major attention in year 3 Physics Analysis over
  Grids

12
Analysis Requirements Prototyping
Remote PROOF Cluster
Local PC
stdout/obj
root
ana.C
proof
.root
TFile
node1
ana.C
proof
.root
GRID as black box
TNetFile
node2
proof
TFile
.root
Select and extract data from grid, Analyze using
local resources without Further interaction with
grid
node3
proof
TFile
.root
node4
13
An Issue for all..
Maintenance and Support

• PPDG statement is that Maintenance, Operations
  and Support is outside the Project Scope.
• Software Maintainenance and support belongs with
  each Development group ie within the Experiment
  infrastructure or with the Computer Science
  Project.
• Issue of longevity of Computer Science Groups an
  important issue which PPDG helps lobby for.
• Operations is the responsibility of each
  experiment or application group - outside of the
  PPDG project.

14
PPDG prospects - current and future

• Provides natural path through well established
  and recognised computer science groups for
  deliverables to be disseminated and used by the
  wider community.
• Provides natural forum for common technology
  components and capabilities to be discussed,
  developed and shared across experiments and
  applications.
• As we move into the final year of the SciDAC
  Pilot we see PPDG Phase III (phase I- NGI phase
  II-SciDAC) necessary to develop the higher layers
  of the application and production GRID and
  Application services
• HENP Data Analysis include Interactive
  capabilities, support for Physics Objects and
  Methods etc.
• Multi-level Security Authorization and Accounting
• Data, Meta-Data and Code Provenance and
  Management.
• Robustness and Fault tolerant Services
• Adoption and Testing of Emerging Technologies
  e.g. Web services.

15
Observations

• Running experiments progress are impacted by the
  need to support data taking and ongoing data
  analysis.
• CS groups get split 6 ways - which makes the
  targetted effort for each experiment small.
• The steering committee has endorsed increased
  collaboration with EDG. Several experiments are
  prototyping/integrating with EDG releases.
• The steering committee has added direct
  collaborations with LBNL-DMF (performance
  monitoring) SLAC-IEPM (netowork monitoring).
• PPDG benefits greatly from leveraged effort from
  all groups.
• Collaboration across the multi-dimensional
  project matrix involves measureable effort.
• Structured Virtual Organization management and
  services, Site/Fabric-Grid interfacing, End to
  End error handlng and Troubleshooting, are
  identified as important issues which require
  increased effort now.
• PPDG goals help and hinder developing final grid
  solutions early results prove the technology but
  one has to address many/all details at once.

16
Trillium

• Name for the collaboration between PPDG, iVDGL
  and GriPhyN. It allows us to
• Make contributions between and across these
  projects without stress of which project is
  doing what
• Work as a unified and consistent entity with the
  EU and other Grid projects.
• Have low-overhead collaborative projects e.g for
  Security, Newsletter, Production Grids and
  Demonstrations.
• Naturally allow the deployment projects to take
  advantage of research projects as technologies
  emerge. E.g. Virtual data in PPDG application
  grids.
• Remembering that
• Experiments and CS groups dont completely
  overlap. Time scales are different. The focus
  varies.

17
e.g. SC2002 Demonstrations
18
Trillium Prospects?

• Continued forum to help the US HEP Grid projects
  form a common view and strategies, lobbying
  group, joint meetings and work.
• Provide a natural collaboration across DOE and
  NSF projects. Is this of benefit to the funding
  agencies?
• might hope for some endorsement and backing of
  the approaches for the future....

19
High Energy Physics Intergrid Coordination

• Open to all participating in HENP grids.
  Participants from EU (specially EDG and DataTAG),
  Asia and US grid projects(Trillium), national
  grid projects (specially GridPP) and experiment
  grid projects (specially LCG)
• Started at the first Global Grid Forum in
  Amsterdam. Commitment to standards shown by
  decision to co-locate all HICB Meetings with
  GGFs.
• Has a strong international focus on policies,
  procedures and technologies, cross-project and
  cross-continent joint projects.
• Past year has shown it to be beneficial as a
  non-competitive forum for discussion and building
  consensus on strategic issues. E.g. Understanding
  of licencing for EDG and US grid middleware
• Joint Technical board (JTB) provides definition,
  oversight and tracking of joint projects -
  coordinators of various grid projects. Co-chairs
  are Ian Bird - LCG Grid Deployment board area
  manager and myself.

20
Grid Laboratory Uniform Environment (GLUE)

• Initial Goals defined in Feb 2002 by JTB. to
  address interoperability from the bottom up
  between application and project grid domains.
• While identical middleware is the shortest path
  to compatability, existing situation has
  different environments and deployments
• Any Deliverables (code, configuration,
  documents) to be deployed and supported through
  EDG and VDT release structure.Effort mainly from
  DataTAG and LCG in EU and iVDGL, VDT and PPDG in
  US.
• Technology Demonstrate that each service will
  interoperate.Provide support for cross-continent
  execution of applications. Identify any lacks and
  incompatabilities early.
• SociologyLearn to work across projects and
  boundaries without explicit mandate or authority
  but for the longer term good of the
  whole.Intent to rapidly expand from 2 continents
  to Global.
• Strategy interoperability as part of the
  ongoing culture of HENP grid projects. Provide
  input to standards and protocols.

21
Glue and Interoperability status and deliverables

• Phase I - 60-80 complete in fall 2002
• GLUE Schema - Common resource information
  definitions presentation across Globus, EDG and
  VDT.
• Support for cross-domain job submission and
  execution.
• Cross-trust of Certificate Authorities,
  interoperability of services for authenticaion
  and authorization based on gridmapfiles.
• Common storage management interface and
  interoperable implementations in the US SRM V1.
• Phase II just getting underway - schedules range
  through 2003
• Storage Management standards V2.
• Release and installation certification and
  testing
• Software meta-packaging, distribution, and
  configuration
• Data Movement - currently handled by the
  experiments at the application level

22
(No Transcript)
23
What coordination takes.. people involved in 2002
(none full time)
24
Plans for HICB, JTB

• Positioning to more actively collaborate with
  Asia.
• GLUE will continue only as long as participants
  and endorsers deem it a useful vehicle.
• Usually surprising new areas of fruitful
  discussion at the HICB meetings.
• JTB becoming a useful umbrella for joint projects
  with in place means for endorsement and buy in
  (manpower always a touchy issue of course).
• As grid projects come and go a loosely defined
  and self-selected forum will continue to be
  useful.
• Co-location with GGF is appropriate as ultimate
  success of the grid depends on open and complete
  standards, principles and policies.

25
In Summary - a PPDG view

• It is useful to have a joint Collaboration
  between experiments already storing and analysing
  up to a Petabyte of data and the LHC experiments
  that are developing and ramping up production and
  preparing for analysis.
• Contacts between the Experiments and PPDG relies
  on the commitment, energy and constant attention
  of the Team Leads..
• Success relies on having sufficient Leadership
  and Management involvement.
• Compromise and acceptance of small steps rather
  than large deliverables seems to give better
  results.
• This brings a danger of postponing or having
  insufficient mass and energy to address the
  difficult or effort intensive needs.
• Inclusive and non-competitive loose coordination
  peer groups are a help.