|
ACS Home Related Links:
Advanced Campus Services |
Minutes of Meeting, June 24, 2003 Present: Xiaochun He (Physics & Astronomy), Jan Isley (Physics & Astronomy), Joseph Liang (Computer Information Systems), Gobinda C. Mishra (Physics & Astronomy), Yi Pan (Computer Science - Yamacraw), Chris Shaw (GVU Center, Georgia Tech), Vijay Vaishnavi (Computer Information Systems), Art Vandenberg (IS&T, Advanced Campus Services), Xiaorong Wang (Physics & Astronomy) Handouts : Minutes of Meeting, May 13 2003 ; Interim Status Report on 7 Action Items ; Specific REU Positions Requested ; Summary on network speed Dr. He's Lab June 24, 2003 Purpose : Check Status Items, Continue discussion on GRID application & strategy 1. NSF REU funding : We have received funding for 2 NSF REU Supplements ($6,000 and $820 travel each) as subaward to SURA NSF Contract #ANI-0123937 entitled "Supporting Research and Collaboration through Integrated Middleware" (NSF Middleware Initiative Integration Testbed Program). Scope of work: cataloging of research application suitable to GRID; assisting in deployment of NMI GRID components; participation in status meetings; contribution to reports, papers, communications. ACTION ITEM : Art will coordinate identification and selection of candidates. 2. Status of Action Items May 13 : updated status as follows: i) Investigate NMI security - Harrison, Vandenberg, et alia - Need to plan for Certificate Authority solution; possible research survey on number of certs deployed in University System? ii) Sun Academic Equipment Grant platforms on GRID - Bolet, Vandenberg - 4 Sun Netra X1s on a local GRID. Ready to issue test accounts. iii) Implement Network Weather Service for our sites - Bolet, Vandenberg, et al - Victor working with Charles Hollingsworth. During today's meeting we agreed on a plan: Move machine that Xiaorong is using for simulation analysis out of lab and connect direct to GSU net core (so bypassing lab network 10mb switched connection, idea being to find performance bottleneck.) Xiaorong will then run analysis, using AFS mounts of libraries at BNL, and see if analysis runs are sustained (from inside lab analysis runs often are killed unexpectedly.) Good performance would enable accessing distributed file catalogs (cf. FROG? grid application - need ref .) ACTION ITEM : Move server, test direct connection to GSU core. Xiaorong, Victor, Charles. iv) Test MPICH-G2 - Harrison - [As of July 1] Victor is working on install of MPICH-G2. Encountered install error - have received response from NMI Support, will continue install. v) Interoperation of PPDG and NMI certificates - Bolet, Vandenberg, Harrison, He - See comments of item i) above. vi) Investigate additional equipment grants (Sun, Dell, Apple, IBM?) - Vandenberg, et al. - No info at this date vii) Applications for a GRID - all participants - Joseph & Victor porting Genetic Algorithm implementation of Self-Organizing Map application. See 3. below. 3. Applying SOM code to the GRID - report on work : Joseph has used CONDOR-NT to submit 4 job runs of Genetic Algorithm to a GRID of three Windows 2000 and XP machines (he submits job from a 4 th machine). CONDOR-NT submits jobs to 3 machines, then submit 4 th job to one of the 3 when it finishes. Advantage of GRID: 4 jobs that take about 80 minutes if run on one machine, run in less time when run across 3 machines. Can also submit a job to one of GRID mounted Sun Netra X1 machines. Performance is quite different. Davide Gaetano, IS&T, noted that an Intel architecture is going to be faster on computation than the SUN X1, which is a low end machine intended more for database, directory use than computation. We'd like to "parallelize" the Genetic Algorithm. though group discussed a number of challenges. Generally, Message Passing Interface (MPI) solutions are expensive in communication overhead between the multiple processors. Still, an MPI solution might be a good learning/research task. 4. Issues with Certificates (Globus CA going away.) - Announcement that GLOBUS CA certificate was being discontinued underlined fact that the GLOBUS certs were "developmental" only, and that we'll need a CA solution for our GRID. Ongoing. 5. Discussion and New Action Item review - The group continued discussing GRID solutions in context of Dr. He's muon detector project. A muon event analysis is comprised of maybe a million discrete events. Clearly, analysis on single machine can takes weeks. Using 100s of GRID based machines could bring results in 10 minutes, say. We sketched out a plan for a GRID that incorporates Dr. He's muon detectors and simulation for high schools. As Dr. He explained the issues that impact muon and which need to be accounted for in simulations, he noted weather and geomagnetic factors. Dr. Shaw suggested adding weather detectors and simulation to the GRID. Art suggested that videoconferencing nodes (videoconferencing is another area of the NMI Testbed work) would be a component of management and learning. Dr. Pan advised that approaching the GRID proposal as a more general, multi-purpose project would enhance its chances with NSF or other agencies, and certainly including mentoring of students and teachers in K-12 is a desirable outcome and would address "broad impact" criteria of NSF.. Art noted that SURA (Southeastern Universities Research Association) NMI Testbed participants were discussing a multi-state GRID for testing. We agreed that planning a multi-purpose GRID would be a good way to focus our work. ACTION ITEM: Art will draft GRID Grant proposal and look at potential NSF programs. ACTION ITEM: Jan will provide summary of research (Quarknet, Simulations, Muon Scope) ACTION ITEM: Investigate grant opportunities - All OPEN ACTION ITEMS : (italics are new items)
Next Meeting : Friday July 18, 2003, 1:30-3:00pm, Commerce Building 1103. |
Last Updated: March 2, 2006