|
ACS Home Related Links:
Advanced Campus Services |
NMI Integration Testbed Program
On the Call: UABJill Gemmill, Pravin R Joshi , John-Paul Robinson, David Shealey UAHSandi Redman UARK (U. Arkansas )Amy Apon Educause/Internet2Ann West GMU ( George Mason University )Phil Yang GSUVictor Bolet, Nicole Geiger, Seema Metikurke, Art Vandenberg, UMichCharles Antonelli, Jim Irrer SURAMary Fran Yafchak, Kate Barzee TACCAshok Adiga SC (U. South Carolina )Kirk Cameron, John Rose USCShelley Henderson TULANEJed Diem, Rene Salmon UVAMarty Humphrey, Jim Jokl
Purpose of meeting - Review status of NMI Testbed GRID; Internet2 demonstration prep; introduction of additional sites joining NMI Testbed Grid; discussion.
Summary of Discussion
Internet2 Demo & Scenarios - We had about 15 minutes discussion on demonstration scenarios. Demonstration description available at: http://events.internet2.edu/2004/fall-mm/demos.html#nmi (see prior minutes as well.) We agreed that we'd schedule the 5 sites such that during each of the 3 afternoon or morning demo times, we'd rotate a "featured site on screen" for 45 minute period, during which others could still demo via laptops. Art will circulate a draft schedule for comment/adjustment. We also agreed that we'd each prepare a Poster of our key points, so that there would be take aways available (or maybe posted where they could be accessed by those interested?) Scenario for the demonstration of the BridgeCA was discussed. Art will circulate an initial scenario to the current cross certified sites (UAB, TACC, USC) and UVA, as a start on refinement of the scenario.
ACTION ITEM : prepare 1 st draft of poster by September 10, send to Art and Mary Fran. ACTION ITEM : draft scenario of BridgeCA.
New sites, SURA Invitation to Participate - In response to July 26, 2004 SURA Invitation to Participate email , we have several new sites joining our Testbed. Each of the sites briefed us on their interest and potential contribution to the overall resources of our NMI Testbed Grid:
- Amy W. Apon , Ph.D., Associate Professor, Computer Science & Computer Engineering, University of Arkansas : "We at the University of Arkansas are interested in participating in the NMI Testbed Grid project. We have a medium-sized cluster [32 dual processor nodes, acquired via NSF MRI grant] that we are interested in adding to the grid.and have several research interests in it." Apon teaches cluster and grid computing, wants some hands on. Very interested in Shibboleth; PI for Educause "Extending the Reach" outreach grant for Great Plains . 3 or 4 application areas of interest at U. Arkansas- computational physics, DNA computing, wind tunnel modeling, geosciences. Shibboleth project would have nice fit with ideas for Shibbolized SURA Sector CA .
- Kirk W. Cameron , Ph.D., Assistant Professor, Computer Science and Engineering, John Rose , University of South Carolina : "We have established a computational grid and portal here at the University of South Carolina for Bioinformatics problems and computer science research. We would be interested in participating and would like to open a dialogue." John Rose, Co-PI on SC Grid also joined call. Have local grid for experimental purposes; BLAST, amino acid analysis; 515 viral genomes; SGI Ultrix (about $70-$100K of machine).
- Jed Diem , Ph.D., Professor, Mathematics, Rene Salmon , Senior Systems Analyst, Tulane : Tulane's Center for Computational Science ( http://www.ccs.tulane.edu / ) has AMD Opteron clusters (2 dual nodes; 7 dual nodes; 2 dual nodes) being deployed with Globus 2.4 and is looking for applications, especially computational fluid dynamics. TACC has interest in computational fluid dynamics and Ashok will find a cfd program that Tulane can use (ACTION ITEM) . Tulane worked with NMI Support to get beta release of Globus for 64 bit Opteron. Still trying to work a problem with running jobs across clusters - seems only 1 processor in each cluster engages. Will offer "crash" [the host name] a 2 node, dual processor AMD to the resources of the NMI Testbed. Tulane also working with 2 node SGI Origin 300.
- Phil Yang , Ph.D., George Mason University (see: http://art12.gsu.edu:8080/grid_cat/index5.jsp and Search Last Name=Yang): April 23 email to Mary Fran: Has project titled "VA-MAGIC" ( http://philler.scs.gmu.edu/vaccess/ ) sponsored by NASA. Building geospatial information services, such as ( http://env.scs.gmu.edu/fairfax/ ). Some of these geospatial information services are computing intensive... For instance, data search and data mining engine takes time - if implemented on grid may improve search time. Looking to adopt Middleware based Grid technology to provide more stable, reliable, and fast services to the users. Have deployed SRB from San Diego Supercomputer Center at UCSD ( http://www.npaci.edu/DICE/SRB/ ) and are in the process of adopting Globus. Could contribute 15 nodes, data searching portal; part of UCAR Unidata consortium ( http://www.unidata.ucar.edu/ ) of weather radar data (4 terabytes of datasets). Possibly interested in UAH contact? Sandi Redman notes UAH also has data search, data ordering and data pool interests.
- Jim Siedow , Vice Provost Research Duke University (not on call, but this info from earlier conference call with Mary Fran and Art). Recognizing the potential of Grids, wants to engage Duke. Jim sought a "key insight" and "activity" that would be a way to move forward. After some discussion, Mary Fran suggested grids need to have IT technical staff to champion and implement the grid and demonstrate capabilities, much as several NMI Testbed Sites are doing. Jim concurs that getting hands on opportunity is important to researchers and a good direction for Duke. Will identify an "IT staff along with a pioneering researcher" team at Duke.
Summary of Action Items:
ACTION ITEM (I2 Demo Sites) : prepare 1 st draft of poster by September 10, send to Art and Mary Fran.
ACTION ITEM (Art and UVA, UAB, TACC, USC) : draft scenario of BridgeCA. (see Appendix)
ACTION ITEM (GSU/USC): Run genome alignment application at USC. Deadline: By June 24? Jim Cottilier reported Sept 7, 2004 that request for HPC resources okay if we will send details and schedule (total wall clock hours expected) for proposed runs. In process.
ACTION ITEM (GSU): Implement OGCE portal, Ganglia. Deadline: By June 24? Rebuilding on Linux server; planned deploy of initial version Friday Sept. 10.
ACTION ITEM (GSU/UAB/USC): Run genome alignment using MPICH-G2. In process. Nova currently evaluating results of several runs. (MPICH-G2 even improves results? Hmmm?) Also working on set of 90 genomes of 80,000 to 1,000,000 length - planned for USC HPC Linux cluster.
ACTION ITEM (ALL): Policy Documentation. Collect any available policy statements from NMI Testbed Sit es as reference resources. (see for instance: a) Grid 2003 User Registration And VO Management Policy listed at http://www.ivdgl.org/grid3/userinfo/ under " General Grid3 user registration Current User Registration document "; b) HPCC Allocation Policy for USC High Performance Computing and Communications, Research Computing Facility listed at http://www.usc.edu/hpcc/ ; c) UABGrid, About at http://lab.ac.uab.edu/uabgrid/title/About .)
ACTION ITEM (Ashok Adiga): Ashok will find a computational fluid dynamics program that Tulane can use.
ACTION ITEM (ALL): Documentation. Ongoing.
NOTE: see http://www.gsu.edu/~wwwacs/GRID_Group/NMI.html for previous minutes of NMI Testbed Grid.
NEXT CALL : Thursday September 23 , 2004 4:00-5:00pm EDT - last call prior to Internet2 meeting
Appendix
September 8, 2004 Suggested scenario, Version 1.0 To demonstrate BridgeCA.
NOTE 1 : This needs to be expanded, tested, and verified by UVA, UAB, TACC, and USC. (Nova Ahmed, GSU, can be resource if running genome alignment).
NOTE 2 : TACC, UAB used as example . While actual scenario may have different sites/roles, this version is based on the fact that the Genome Alignment algorithm has already been tested and run on the UAB Beowulf cluster (as well as across multiple clusters.) So that at least has part of the problem addressed.
Scenario Title : TACC user runs genome alignment job on UAB resource.
1. User at TACC authenticates to local TACC system
2. Authenticated TACC user does myproxy init to generate grid cert
3. TACC user submits globusrun command to execute job at UAB grid resource Suggest running Genome Alignment application - can be set to execute in several minutes (by varying number and size of genome sequence data), allowing viewing of resource use.
4. UAB grid resource recognizes and accepts TACC certificate
5. TACC user views UAB resource running job (cf. cpu utilization.)
6. TACC user retrieves UAB results (is this gridftp.?)
7. Perhaps viewing of log files will be helpful to show "that it actually did work."
|
Last Updated: March 2, 2006