Open Science Grid Ruth Pordes Fermilab V2 3/27/07 ISGC2007: OSG - - PowerPoint PPT Presentation

3/27/07 ISGC2007: OSG 1

Open Science Grid

Ruth Pordes Fermilab V2

3/27/07 ISGC2007: OSG 2

First of All

• I am in Taipai in spirit today but not able to be there

in person.

• A big thank you to Simon and Vicky for enabling me

to give my talk remotely

• And thank you to Stephen, my husband, is with my 86

year old father who arrived from England today - one

• f the several family reasons that kept me in the US

this week. ISGC is now an annual event where truly the ISGC is now an annual event where truly the whole earth aspects of science on the whole earth aspects of science on the distributed infrastructure is clearly shown. distributed infrastructure is clearly shown.

3/27/07 ISGC2007: OSG 3

Outline

• Status of OSG
• Architectural Model
• OSG & LHC
• Some Sample Activities

Note: OSG is International in scope International in scope with its direct goal to deliver a Virtual Distributed Facility Virtual Distributed Facility in the US.

3/27/07 ISGC2007: OSG 4

What is the Open Science Grid?

• People and Organizations

People and Organizations as members & partners of a Consortium Consortium with Common Goals Common Goals to support Computer Based Research over wide and local areas of distribution.

 Includes Facilities, Experiments, Computer Scientists, Grid organizations etc.

− − Global in Reach: Currently include Taiwan, South Africa, South Global in Reach: Currently include Taiwan, South Africa, South America, Australia, UK America, Australia, UK

• Contributor

Contributor to the vision of Cyberinfrastructure Cyberinfrastructure “.. a cultural community that .. enables distributed knowledge communities .. collaborate and communicate across disciplines, distances and cultures.”

 Dr Arden Bement, Director of National Science Foundation in “Cyberinfrastructure Vision for 21st Century Science.

− − Global in Scope: NSF actively discussing how to increasingly Global in Scope: NSF actively discussing how to increasingly sponsor international partnerships. sponsor international partnerships.

• “Each participant retains its separate legal status and the consortium's control
• ver each participant is generally limited to activities involving the joint

endeavor”

 From Wikipedia definition of Consortium

3/27/07 ISGC2007: OSG 5

And the OSG Project

• The OSG Project has been awarded ~$6m/year for 5 years to

deliver to the goals of the Consortium.

• Support across 6 program offices of the DOE Office of Science

SciDAC-2 program and the National Science Foundation. − − International components for International Science Grid This International components for International Science Grid This Week online newsletter, and outreach and educational Week online newsletter, and outreach and educational activities (currently to Scandinavia - IceCube - and South activities (currently to Scandinavia - IceCube - and South Africa - ATLAS). Africa - ATLAS).

• Support for 30 OSG staff at 16 DOE Laboratories and Universities:

University of Iowa University of Chicago/Argonne National Laboratory University of Florida University of Wisconsin, Madison University of California, San Diego Stanford Linear Accelerator Center Rennaisance Computing Institute Lawrence Berkeley National Laboratory Indiana University Fermi National Accelerator Laboratory Cornell University Columbia University California Institute of Techology Brookhaven National Laboratory Boston University

3/27/07 ISGC2007: OSG 6

OSG Organization

3/27/07 ISGC2007: OSG 7

OSG Project Scope

• independent of national boundaries
• Maintain and Operate a Secure Production Quality Common

Secure Production Quality Common Distributed Infrastructure Distributed Infrastructure providing access to & sharing of access to & sharing of OSG Consortium members and partners computational and storage computational and storage resources resources over production and research networks.

• A reference distribution of common technologies for high

A reference distribution of common technologies for high throughput distributed computing throughput distributed computing -- the Virtual Data Toolkit (VDT) Virtual Data Toolkit (VDT)

• whose components are requested by OSG members and partners

and which is built, tested, integrated, packaged and supported.

− External projects that develop the software and/or we rely on and/or we deliver to are part of the OSG management structure.

• Support for Distributed Communities

Support for Distributed Communities -- of resource owners, science and research users, educators and students, technology developers -

• of both existing and new OSG members to

to use and evolve the use and evolve the system. system.

− Including a multi-site integration testbed for bringing new software and services into production use.

3/27/07 ISGC2007: OSG 8

External Projects:

• Center for Enabling Distributed

Petascale Science (CEDPS)

• Community Driven Improvement
• f Globus Software

(CDIGS)/Globus

• Condor
• dCache
• Data Intensive Science

University Network(DISUN)

• Disk Resource Manager (DRM)
• ESNET
• Internet2
• LIGO Physics and the

Information Frontier

• OSG Accounting
• OSG Privilege/Authorization
• UltraLight/advanced networks
• U.S. LHC + US representation to

the WLCG

• Autralian Partnerships for Advanced

Computing (APAC)

• Data Intensive Science University

Network (DISUN)

• Enabling Grids for E-SciencE (EGEE)
• Grid Laboratory of Wisconsin (GLOW)
• Grid Operations Center at Indiana

University

• Grid Research and Education Group at

Iowa (GROW)

• Nordic Data Grid Facility (NDGF)
• Northwest Indiana Computational Grid
• New York State Grid (NYSGrid)
• TeraGrid
• Texas Internet Grid for Research and

Education (TIGRE)

• TWGrid (from Academica Sinica Grid

Computing)

• Worldwide LHC Computing Grid

Collaboration (WLCG)

Partners:

3/27/07 ISGC2007: OSG 9

Science Community stakeholders

• global in reach

3/27/07 ISGC2007: OSG 10

100 Resources across production & integration infrastructures !ncrease in ~15 in last 6 months 27 Virtual Organizations (+ 3 operations VOs) 25% non-physics. ~20,000 cores (from 30 to 4000 cores per cluster) ~6 PB accessible Tapes ~4 PB Shared Disk Sustaining through OSG submissions: Measuring ~180K CPUhours/day. ~Factor of 50% more (being measured) in last 6 months. Using production & research networks

OSG Today

3/27/07 ISGC2007: OSG 11

OSG Architectural Model (1) Importance of Virtual Organization (VO)

Courtesy: Frank Wuerthwein

3/27/07 ISGC2007: OSG 12

OSG Architectural Model (2) Vertical and Horizontal Integration

Infrastructure Applications

VO Middleware

Core grid technology distributions:

Condor, Globus, Myproxy: shared with TeraGrid and

• thers

Virtual Data Toolkit (VDT)

core technologies + software needed by stakeholders:many components shared with EGEE

OSG Release Cache: OSG specific configurations, utilities etc.

HEP Data and workflow management etc

Biology Portals, databases etc

User Science Codes and Interfaces Existing Farms, Storage, Networks

Astrophysics Data replication etc

3/27/07 ISGC2007: OSG 13

OSG Architectural Model (2) Vertical and Horizontal Integration

Infrastructure Applications

VO Middleware

Core grid technology distributions:

Condor, Globus, Myproxy: shared with TeraGrid and

• thers

Virtual Data Toolkit (VDT)

core technologies + software needed by stakeholders:many components shared with EGEE

OSG Release Cache: OSG specific configurations, utilities etc.

HEP Data and workflow management etc

Biology Portals, databases etc

User Science Codes and Interfaces Existing Farms, Storage, Networks

Astrophysics Data replication etc

3/27/07 ISGC2007: OSG 14

OSG Architectural Model (2) Vertical and Horizontal Integration

Infrastructure Applications

VO Middleware

Core grid technology distributions:

Condor, Globus, Myproxy: shared with TeraGrid and

• thers

Virtual Data Toolkit (VDT)

core technologies + software needed by stakeholders:many components shared with EGEE

OSG Release Cache: OSG specific configurations, utilities etc.

HEP Data and workflow management etc

Biology Portals, databases etc

User Science Codes and Interfaces Existing Farms, Storage, Networks

Astrophysics Data replication etc

• ≈

3/27/07 ISGC2007: OSG 15

OSG Architectural Model (3) Equivalence of VOs and Resources

• Site (Resource) and VO Administrators have equivalent

equivalent security, testing, monitoring, Roles. security, testing, monitoring, Roles.

• Site (Resource) Owners and VO Managers have

equivalent trust, policy, accounting, Responsibilities & equivalent trust, policy, accounting, Responsibilities & Authorities. Authorities.

• Users share

Users share in VO and Site (Resource) Services.

3/27/07 ISGC2007: OSG 16

OSG Architectural Model (4) Overlapping & Interfaced Environments

Equivalence of Campus, Community and National Infrastructures with

• verlapping Environments.

Heterogeneity of Interfaces and Services - minimally evidenced by multiplicity

• f versions.

3/27/07 ISGC2007: OSG 17

OSG delivers to ATLAS, CMS and the WLCG

• Deliverables and milestones agreed to by OSG Executive Board

(project) and Council (consortium).

• OSG is the distributed facility for ATLAS and CMS in the US.
• OSG reports to the WLCG and meets milestones defined by the

experiments for that Collaboration.

• OSG reports to US ATLAS and US CMS software and computing

and meets deliverables defined by those projects.

2007: Deploy SRM V2.2 (dCache and DRM) on all US LHC Storage sites. Deliver to throughput requirements for Data - Tier-1s to Tier-2s (see Don’s talk)- and Jobs (>100K per day). Meet Availability and Success metrics.

3/27/07 ISGC2007: OSG 18

Sample activities

• Maintain and Operate a Secure Production Quality

Secure Production Quality Common Distributed Infrastructure Common Distributed Infrastructure providing access access to & sharing of to & sharing of OSG Consortium members and partners computational and storage resources computational and storage resources over production and research networks.

3/27/07 ISGC2007: OSG 19

production qualities: Security

• Well -- Don and Bob are there to tell you all about it..

Increased need for and scope of Agreements Awareness training and tracking ~50 page Risk Assessment ~30 item Security Plan Comprehensive set of Examination Audits and Reviews. TeraGrid testing Shibboleth -> X509. OSG watching until integrations done by NCSA, EGEE and end-to-end deployments on TeraGrid work.

3/27/07 ISGC2007: OSG 20

production qualities: Availability & Accounting

• Project starting now to acquire (repackage site_verify and SAM

tests) Validation Probes Validation Probes, add framework for Site Control of Site Control of Execution Execution, report to WLCG availability repository: June 2007.

− Yes, working with/as part of of the WLCG Monitoring Group,

• CPU accounting

CPU accounting now at 38 production sites. Already used to spot problems (lack of use, too much use?)

− Also measure of measure of “ “sharing sharing” ”: : use of CPU by VOs other than those that Own the resource. − Interfacing form “pull-/glide-in” mode jobs.

• dCache transport accounting

dCache transport accounting available in production - only just starting to gather information.

− Expect to also collect GridFTP storage elements transport accounting.

3/27/07 ISGC2007: OSG 21

sharing of CPU, Data, Storage

• CPU sharing managed by configuration of Batch Queues and

mapping of identification certificate role attribute to account UID.

− Preemption allowed but causes problems for many applications. − Will soon need agreements between Sites and VOs. − OSG VO has allocations for the Consortium (e.g. LBNL) and for these decides on priorities and expectations.

• Data sharing managed by mapping of identification certificate

role attribute to access control and root path to data.

− Still no good test of interoperation with EGEE.

• Storage sharing currently through static allocations and file

system allocations. In the future will support reservation and release through SRM.

3/27/07 ISGC2007: OSG 22

sample activities: Software

• A reference distribution of common technologies for

A reference distribution of common technologies for high throughput distributed computing high throughput distributed computing -- the Virtual Virtual Data Toolkit (VDT) Data Toolkit (VDT) -- whose components are requested by OSG members and partners and which is built, tested, integrated, packaged and supported.

− External projects that develop the software and/or we rely on and/or we deliver to are part of the OSG management structure.

3/27/07 ISGC2007: OSG 23

The VDT

• OSG VDT supports TeraGrid, EGEE, OSG, APAC, NWICG

TeraGrid, EGEE, OSG, APAC, NWICG and others.

• VDT 1.6.0 and OSG 0.6.0 released

VDT 1.6.0 and OSG 0.6.0 released and being installed on OSG.

− Includes SRM/dCache distribution, now installed at ~5 sites. − EGEE CEMON used to publish GLUE information to OSG-BDII. − Class-Ad based resource selector available. − Integration and provisioning of OSG release took 2-3 months.

• VDT 1.8.0 for release in June

VDT 1.8.0 for release in June list (OSG 0.8.0 for August) and priorities in discussion.

− Build subset collections for EGEE − Add VOMRS, glEXEC, SRM V2.2

• Other Activities

− Discussions underway to build subset of VDT under ROCKS for PRAGMA distribution. − Looking to make better use of NMI Build and Test (Metronome) and commonality with ETICS.

3/27/07 ISGC2007: OSG 24

Software: Development

• OSG works closely with many software development

groups:

− Accounting, Authz are external projects. − Condor, Globus, dCache are external projects. − EGEE develops software that OSG is interested in. − Etc.

• OSG provides a valuable testbed for new CS software

methods and technologies.

• Portals and User interfaces coming.

3/27/07 ISGC2007: OSG 25

sample activities: Communities

• Support for Distributed Communities

Support for Distributed Communities -- of resource

• wners, science and research users, educators and

students, technology developers -- of both existing and new OSG members to to use and evolve the system. use and evolve the system.

− Including a multi-site integration testbed for bringing new software and services into production use.

3/27/07 ISGC2007: OSG 26

communities: Engagement, Extensions

• RENCI working with Rosetta application group which has run

production on OSG.

• Several biology communities at UCSD All hands meeting starting

to test their legacy applications on OSG. No large production runs recently though.

• Popular molecular dynamics code - CHARMM - successfully running

under ATLAS PANDA framework.

• Site logging and reporting based on syslog-ng starting tests on

ITB for deliverable from LBL in late summer.

− Will also look at use cases where Globus Auditing service might benefit .

• WS-GRAM in test on ITB and by several small communities.

− Resulting in enhancements to prevent overload and better error messages.

3/27/07 ISGC2007: OSG 27

communities: Grid Interoperability Now

• OSG participating in new GIN charter.
• PRAGMA-OSG tests continue with applications running

from both sides.

• Flexible and extensible build and test might help?

3/27/07 ISGC2007: OSG 28

communities: Joint CI Days

• Cross-campus cyberinfrastructure

Cross-campus cyberinfrastructure will need increased cross-campus communication and working together -- change that comes slowly.

• University hosts a CI Day

University hosts a CI Day locally to bring together the local communities - CIO, IT departments, research CIO, IT departments, research departments, faculty and educators departments, faculty and educators - to discuss and define the needs and ideas for cross-campus cyberinfrastructures.

• OSG, Internet2, TeraGrid, Educause

OSG, Internet2, TeraGrid, Educause help organize and attend the meeting, and together work with the campus organizations to present what is available today and to discuss the university’s needs.

3/27/07 ISGC2007: OSG 29

communities: Grid Schools

• OSG plans hands-on grid schools 3 times a year.
• First one in Chicago Mar 24-25th (weekend) with 60

attendees.

• OSG co-sponsoring International Grid School 2007.
• First training of Site Administrators at US LHC Tier-

3 worksop co-located with OSG all-hands meeting.

− Planning more site administrators training.

3/27/07 ISGC2007: OSG 30

In Summary

• OSG established as Project in support of an open,

inclusive and international Consortium.

• OSG plans to deliver to LHC needs and schedule.
• OSG implementation architecture guided by it’s

architectural model. Thank you!