|
ACS Home Related Links:
Advanced Campus Services |
NMI Testbed GRID On the Call: UABJason Lynn UABSandi Redman GSUArt Vandenberg Internet2Steve Olshansky UMichVictor Wong TACCAshok Adiga Purpose of meeting - Review status of NMI Testbed GRID. Focus was post-Internet2 Fall results workshop and potential Internet2 Session and/or demo. Discussion of action itemsACTION ITEM (ALL): provide initial script for potential demo and/or post-Internet2 Fall results workshop. This call centered on upcoming Internet2 Fall meeting. The NMI Integration Testbed will have a Thursday PM/Friday AM follow on Results Workshop, wrapping 3 years (May 2002-Dec 2004 with no-cost extension). [Note: we subsequently submitted on June 7 session and demonstration proposals to Internet2 Fall meeting itself] OGCE Portal - seems to be the way to go. TACC and UMICH are developing/implementing as part of their 2003 NSF Open Grid Computing Environment (see: http://www.nsf.gov/od/lpa/news/03/pr0399.htm ) award. UAB is building OGCE portal. Zach Garner, John-Paul Robinson, and Pravin Joshi of UAB have been helpful GSU's portal plans. GSU will drop GridPort and work on OGCE with Ganglia for monitoring resources (see: http://ganglia.sourceforge.net/ ). Pravin is working on getting MPICH-G2 installed UMich has portal with access via KX.509. It's been tested from Windows, Linux, and Mac platforms; access from Hong Kong. Victor Wong is thinking about how would this play/demo in the testbed environment. UAH portal effort is underway with Dennis Ganon. Metascheduling - Both TACC and UMich are working on portal solutions for meta-scheduling. Victor Wong noted that the UMich meta-scheduler had a tie-in with Network Weather Service. Certificates - TACC's bridgeCA is being re-worked, tested. UAH is looking to cross-certify with the UVa bridgeCA. We agreed that a demo of the bridgeCA would be useful - for instance: can a user with a UAB cert access resources of USC without recourse to using a USC issued cert? Applications - Applications are good way to demonstrate the NMI Testbed Grid. Nova Ahmed has provided UAB with her source and instruction on how to run. UAB will incorporate into demo of UAB portal to UAB faculty. Nova has contacted USC about access to resources (currently in process - demonstrating a Policy Issue aspect of the NMI Testbed Grid: what are the rules? Do the rules vary across sites? We need to monitor this.) Observed need for documentation: How, in fact, does each site handle access to grid resources? What's the baseline that we are working with and what recommendations for NMI Testbed Grid users? NSF REU - TACC and GSU will schedule conference call to discuss TACC NSF REU and collaboration options. Summary of Action Items: ACTION ITEM (ALL): provide initial script for potential demo and/or post-Internet2 Fall results workshop. Consider especially demonstrating access to external resources - which is what the NMI intra-Testbed Grid is about. Deadline: By early June. ACTION ITEM (TACC/GSU): Coordinate on NSF REU resources Deadline: In next two weeks. ACTION ITEM (GSU/USC): Run genome alignment application at USC. Implies getting USC cert and appropriate policy for access. And sets up potential of using USC cert to access UAB genome alignment application (and vice versa.) Deadline: By next call? (June 24) ACTION ITEM (GSU): Implement OGCE portal, Ganglia. Deadline: By next call? (June 24) ACTION ITEM (GSU/UAB): Run genome alignment using MPICH-G2. Deadline: ? ACTION ITEM (ALL): Documentation. Document demo script; document current policy for access to grid resources; document status of portal implementation (providing URL at least). NEXT CALL : Thursday June 24, 2004 4:00-5:00pm EDT ACTION ITEM Response Georgia State Scripts for post-Internet2 Results Workshop (possibly Internet2 Session and/or Demonstration) NOTE: Not validated yet, nor verified with my own Georgia State people, but this is a start of a script. Input, corrections, advise more than welcome. SCRIPT 1. Non-NMI Testbed Portal SSH login Submit job (genome alignment) View results (Patter: "Can be done; limited friendliness; little middleware support; not scalable.") SCRIPT 2. NMI Testbed Portal - A URL accessed Display various resources (minimum Georgia State and one external GSU resource; ideally several sites) Login, MyProxy init (Globus cert for GSU local testing) Submit job to GSU resource (genome alignment) View results (Patter: "Can be done better; more friendly; middleware support for local organization access; scalable.") SCRIPT 3. NMI Testbed Portal - B URL accessed Login, MyProxy init (UAB cert) Submit job to UAB resource (genome alignment) View results (Patter: "Even better; more friendly; middleware support for extra organizational access; AuthN/AuthZ for external resource.") SCRIPT 4. NMI Testbed Portal - C URL accessed Login, MyProxy init (UAB cert) Submit job to USC resource (genome alignment) View results (Patter: "Way better; more friendly; middleware support for cross-organizational access; AuthN/AuthZ scalable beyond organization.") TRUMP CARD? : Genome alignment algorithm finds successively better results. Georgia State 24-cpu shared-memory local system vs. UAB new 8-node, 32-cpu Beowulf cluster vs. USC ???/or alternate platform. That is, application results are improved by access to grid middleware solution: Portal, AuthN/Z and bridgeCA, resource sharing.
|
Last Updated: March 2, 2006