TIGREA TEXAS COLLABORATIVE GRID

Phil SmithSr. Director, TTU HPCC

February 22, 2007

Copyright Philip W. Smith 2007. This work is the intellectual property of the author. Permission is granted for this material to be shared for non-commercial, educational purposes, provided that this copyright statement appears on the reproduced materials and notice is given that the copying is by permission of the author. To disseminate otherwise or to republish requires written permission from the author.

What is a Grid?

• Policies and procedures to enhance shared use

• A reason (or reasons) to collaborate

• Middleware (Globus, VDT, SRM, etc.) to enable work to be done

• Resources Computation Storage Network, etc.

A grid typically has these components:

HiPCAT

Formed in 1999 for its members to share expertise and computing experiences

HiPCAT has 10 University members to date

TIGRE is a project of the HiPCAT consortium

Rice

UT Austin

TTU

TAMU

UH

UT Arlington

UTEP

UTSMC

BCM

UTHSCSA

High Performance Computing Across Texas

HiPCAT Members

TIGRE Project

Texas Internet Grid for Research and Education Created as a HiPCAT project Primary goal is to design and deploy a grid computing

infrastructure that integrates computing, storage, visualization labs, displays, sensors, and instrumentation across Texas

Primary Members:

Provides support for other schools, organizations and enterprises

TIGRE is a Grid construction project to provide a mechanism for Institutional collaborations

TIGRE Grid

InternetInternet

HiPCAT/TIGRE Objectives

To build a nationally recognized infrastructure in advanced computing and networking

To enhance collaborations among Texas Institutions of higher education and Texas industry with national and international in science and engineering projects

To enable Texas Scientists and Engineers in academia and industry to pursue leading edge research in biomedicine, energy and environment, aerospace, materials science, agriculture, and IT

To help educate a highly competitive workforce that will allow Texas to assume a leadership role in the national and global economy

LEARN

Lonestar Education And Research Network LEARN is a cooperative effort of 33 TX Institutions of higher

education

Manages relationships with Internet2, National Lambda Rail, and other regional and commercial ISPs

Targeted to provide Gbit (and higher) network speeds

More information at http://www.tx-learn.org

LEARN Topology

TIGRE and LEARN Funding

TIGRE and LEARN were awarded support from the Texas Enterprise Fund

$2.5 M for TIGRE as a grid software development effort

$7.8M for LEARN networking projects are the strategies for HiPCAT efforts

TIGRE Organization

Consists of a Steering Committee and a Developers Team Two year project plan with activities defined to meet quarterly

milestones Official start date was Dec 1, 2005 Steering Committee sets overall directions (J. Boisseau, Lee

Panetta, Lenart Johnsson, Jan Odegard, Phil S.) Development Team Activities:

• Weekly Telecon, visits with potential users, collection of applications and resource requirements, deploying candidate applications, software component selection, outreach, and Grid middleware support

Software stack defined and already released Currently building a gateway: http://tigreportal.hipcat.net

TIGRE Application Areas

Steering Committee in consultation with researchers and developers has targeted applications to be supported in the following areas:

Bioscience and Medicine

Energy Exploration

Environmental/Air Quality Modeling

Developers are working to implement applications in these areas

TIGRE Resources (http://tigreportal.hipcat.net)

Parallel Computing Resources

Name Institution Department System CPUs Peak

GFlops Memory GBytes

Disk GBytes

Cosmos Texas A&M University

Texas A&M Supercomputing Facility

SGI Altix 128 666 256 4096

Eldorado

University of Houston

Texas Learning and Computation Center

Eldorado Itanium2 Cluster

124 472 256 2232

Lonestar The University of Texas at Austin

Texas Advanced Computing Center

Dell PowerEdge Linux Cluster

5200 55000 10400 94900

Minigar Texas Tech University

High Performance Computing Center

Dell Linux Cluster

32 230 64 70

RTC Rice University

Computer and Information Technology Institute

HP Itanium II Linux Cluster

290 1044 596 7000

Weland Texas Tech University

High Performance Computing Center

AMD Athlon MP 2000+

64 69 64 780

Total: 5838 57481 11636 109078

TIGRE Timeline

Dec 1, 2005

Q1 Q2 Q3 Q4

Q5 Q6 Q7 Q8

–Project plan–Web site–Certificate Authority–Test bed requirements–Driving applications

TIGRE portal –TIGRE Software stack–Distribution Mechanism–Demonstrate one app

Client softwarepackage

Customer Mgmt SysDemo 3 applications

Today

Global scheduler

–Freeze SW features –TIGRE service requirements

–Final Software,–Documentation, –Policies and –Procedures to join TIGRE–Demonstrate–TIGRE Grid at–SC07

Y2

Y1

Nov 30, 2007

Successes

We have ported several applications to TIGRE including:

• UltraScan: Using ultracentrifugation to determine size and geometry of large molecules. Researcher: Borries Demeler at UTHSCSA.

• ENDyne: Quantum Chemical Dynamics. Researcher: Jorge Morales, TTU

• Proth: Number Theory. Researcher: Chris Monico, TTU

• Working on MCNPX and Geant4 for cancer radiotherapy modeling. Researchers: Wayne Newhauser, MDACC and Murali Nair, JACC in collaboration with TTU researchers.

• Pursuing Air Quality Data transfer in collaboration with TTU, TAMU and UH and Petroleum Engineering applications with UT, TTU and Rice faculty members.

Successes (2)

TIGRE has high visibility on a national and International scale

• PTCOG45 (Particle Therapy Cooperative Group) Oct. 7-11 in Houston

• TAGPMA (The Americas Grid project management authority)

• Educause

• SURA and SURA GRID

• I2

• OGF 19, Chapel Hill

Merits of TIGRE Vision

• TIGRE has already spawned new and interdisciplinary collaborations

• A model for higher education resource sharing

• Academic-industrial-government partnerships are possible

• Excellent resource for collaborations within and beyond institutional collaborations

TIGRE Challenges

Hardware

• Fewer free cycles now than three years ago

• Each University has to justify placing a resource on the Grid

• Heterogeneity of hardware

TIGRE Challenges (2)

Applications

• Chicken and Egg principle

• Not all applications are appropriate

• Additional coding and porting required

• Commercial codes (licensing)

• Code compatibility

• Large Data sets (1-2 GBytes)

• Export controls

TIGRE Challenges (3)

Middleware

• Packaging

• Updating

• Web services Follow on funding

It is our observation that, internal (intra-university) collaborations are (sometimes) much harder to foster than external collaborations

Questions?