|
ACS Home Related Links:
Advanced Campus Services |
Status meeting 1. Funding status (15 minutes) NSF REU - recommendation to focus work on Application Catalog NMI Integration Testbed: intra-testbed GRID (2 GRAs?) 2. Email to Dr. Charles Louis, VP Research (10 minutes) 3. Status of Action Items (10 minutes) 4. Draft GRID Proposal (30 minutes) Multi-phase or parallel? Component approach. NMI Intra-Testbed - GSU lead, UAB, UAHuntsville, USC, Mich, UVa. Muon detector Multi-Purpose 5. Discussion and New Action Item review (20 minutes) Next meeting: TBD Interim status report on 7 Action Items from last meeting i) Investigate NMI security (review certs & SSH; test?) - Harrison, Vandenberg, et alia Information from NMI Grid conference call: Globus Certificates not meant for production. Looks like we need to provide Certificate Authority solution. Thinking of a University System wide project: Catalog & Analysis of Cert Usage. GCATT contact re FUNDING? ii) Put Sun Academic Equipment Grant platforms on GRID - Bolet, Vandenberg 4 Sun Netra X1s on a local GRID. Ready to issue test accounts. Note: GRIDs are "not just for parallel processing." Also a means of accessing wider set of resources (you can get to resources that you may not have/own locally) or enabling access to cycles (to run PROGRAM_A four times, you can use 4 machines at once rather than just 1 machine 4 times.) iii) Implement Network Weather Service for our sites - Bolet, Vandenberg, et al Victor contacted Charles Hollingsworth. Has MAC addresses of several Dr. He's machines. Idea is to start with regular network analysis... then compare to NWS. iv) Test MPICH-G2 - Harrison No info at this date v) Interoperation of PPDG and NMI certificates - Bolet, Vandenberg, Harrison, He Given that Globus Certificates are for testing and not recommended for production, interoperation is probably not the issue. Need to understand PPDG certs and their level of security. See comments of first item 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 working on porting Genetic Algorithm implementation of Self-Organizing Map application. Making progress. David, Art, Victor discussed options for rendering application. Need to follow through. . Georgia State UniversityGeorgia State University has been working for nearly 3 years to develop research and enterprise middleware capability and has consistently involved graduate as well as undergraduate students in this work. REU positions in middleware development and integration are proposed in two existing projects: "Muon particle detector GRID for K-12" - Dr. Xiaochun He is working on involving K-12 schools in the physics related to muon particle detectors. Dr. He has designed low-cost muon detector hardware to be deployed in selected K-12 schools in Georgia, creating a distributed network of detectors. The use of GRID technology to manage and access these distributed resources is a desirable outcome. The student assigned to this project would participate with Dr. He in the design and deployment of the muon-detector project in Georgia high schools and contribute to the preparation of a related report on K-12 GRID deployment issues and outcomes. The student would help deploy the Globus Toolkit and Network Weather Service sensor components to manage the distributed nodes of muon detectors. The student would get experience with collaborative research tools, hands-on experience establishing an experimental muon detector grid, and experience in how such a project fosters interest in physics research among K-12 students. The student's work would also contribute to GSU's long term goal for GRID access to the RHIC at Brookhaven and LHC at CERN. "GRID-enabled Applications, including Graphic Rendering with the GSU Digital Aquarium" - GSU is supporting graphic rendering for biology and providing support for students to use visualization as a learning modality. The newly opened Digital Aquarium lab ( http://www.gsu.edu/~wwwdaq/ ) provides students with high-end workstations equipped with multi-media production tools. The ability to bring these high-end resources onto a GRID would provide access of significant benefit to students and faculty. The CIO and the VP of Sponsored Research in GRID services have also agreed to support expansion of GRID technology for "intensive computation". Dr. Harrison is engaged in building GRIDs and works with Chemistry and Biology researchers (cf. Dr. Irene Weber and others) in options for modeling of complex molecular structures. The student on this project would GRID-enable the Digital Aquarium resources and amend related applications to execute effectively over this GRID. The student would also participate in cataloging other applications to be "GRID-enhanced" and evaluate and report on the effectiveness of doing so . The student would specifically investigate GRID development for SERCAT ( http://www.ser.aps.anl.gov/new/index.html ), a distributed crystallography initiative involving several NMI Integration Testbed sites. SUMMARY on network speed Dr. He's lab, June 24, 2003 (per Charles Hollingsworth & Victor Bolet's work) On checking trace routes from Dr. He's lab to BNL, it appears that the connection goes from GSU via ESnet and ESnet provides expected speeds. The limits appear to be due to GSU side, in particular, the 10meg switched connections to Dr. He's labs and, maybe, the DS3 link. We are going on the assumption that Dr. He's application need is to link files at BNL via AFS with reasonable response rates (so as to enable effective compilation of programs at GSU where they will have AFS access to needed libraries.) Recommended that we take one of Dr. He's machines, connect direct to GSU Internet2 core to take advantage of best possible connection (i.e. take his lab connection out of the picture for the moment). Hypothesis is that bypassing the lab's 10meg switch will improve response. If that's the case, then there are a number of options including, improving to link from Dr. He's lab to GSU Internet2 and/or possibly looking at ESnet (DS3 is 45mbps). It is noted that Dr. He's lab's internal network can't be directly connected to GSU Internet2 unless it conform to operational standards. There is a 100mbps port that could enable one of his servers to be connected at that speed to GSU Internet2, and that server, in turn could be connected "on its other side" to the lab network, so that server could mediate between the GSU Internet2 and the internal lab network (but internal net would NOT have direct access to outside.) If the lab port were to be upgraded to GigE speed, that might cost $10,000 or so (depending...). Other options depend on whether this is a "one-time and only for Dr. He's lab" solution or whether there's benefit to other labs and applications. Charles noted that if further throughput requirements were needed, a possibility for improving the ESnet connection (DS3 at 45mbps), might be to go to the 56 Marietta Street location where the ESnet southern node terminates (we'd have to convince ESnet that there is a business case.) Upgrading that could improve speeds while at the same time lower monthly costs for ESnet (and so the upgrade costs might be offset.) But again, upgrading ESnet depends on whether Dr. He's application and/or others can justify the solution. KEY FACTOR : Applications can drive the technology justification. We need to understand and make the appropriate business case. |
Last Updated: March 2, 2006