June 23, 2006

Attending

Micah Beck, Terry Moore (UTK) Surya Pathak, Bobby Brown, Alan Tackett, Paul Sheldon, Larry Dawson (VU)

Meeting Minutes

NSF Workshop funding

next proposal deadline is September 20. CITeam is primarily for education

Status of Library of Congress proposal

pre-proposal submitted but we have heard nothing yet. Assume response should be coming soon. Greenlight somebody soon. Hear about it in the next few weeks.

Brazil Trip issues

Terry talked to Hunter yesterday. Current state of schedule, etc.: the thing he sent us is the general framework, does not include individual meetings. The bioinfomatics folks are very interested, and are making it a high priority to come talk to us. What about collab with VU folks? Will have to find out when we get down there. What are their data logistics issues, volumes of data?

Physics community: Paul has heard from both people, person in Rio wants us to come there. Paul should see if he can send one person to Sao Paulo, and should forward responses he got to the whole group.

New guy working at RMP, very excited about what we are doing. Background in earth sciences, working with San Diego SC center. How can REDDnet and their depot network in Brazil peer together or work together? Interested in working out a plan for that.

Hunter was hoping Alan could physically bring him the Capricorn unit. Hunter is hopefully working that out with Capricorn. How do we get it through customs. Terry points out they have a tie in to Library of Congress. Capricorn is a spin-off from Library of Congress work (Brewster Kail (sp?)).

They have our hotel rooms for us, will be about $55/night. Don't know location is yet.

Martin Swany

Alan talked to Martin Swany, he is interested in "multi-pathing" aspects of REDDnet development. This was his Ph.D. thesis. Martin is an old friend of Terry and Micah. Faculty at University of Delaware. Get a graduate student or so on board? Get involved in UltraLight. He is the only Internet 2 Faculty Fellow. Worked for Jack for awhile.

UltraNet Status

not later than September 30 to light TVA fiber, will run from Oak Ridge to Memphis.

ORNL Status

We do have a compute node (just fixed) and one eonstore (busted). We have two depots at Fermilab to move, one to Michigan, and the other to ORNL.

Bobby will get all of this work scheduled

we have an "in". Interest from John Cobb of TeraGrid. Could probably get some equipment on the TeraGrid.

REDDnet Performance

automatic test framework is now ready and profiling tests have started. see REDDnet wiki page, section of Profile Analysis. We plan to finish profiling then start with basic network tuning. Start with TCP/IP buffer sizes, before starting to use things like FAST. Right now from VU to UFL we are plateuing at 100 Mbs. Not sure what path we are going, Internet 1 or Internet 2. Our I1 connection is 100 Mbs, so that may be limitation. After profiling and network tuning, will start working on parameter sweep to characterize performance. After we finish testing L-Store, will start looking at other SEs such as SRM servers. Alan talked to Michigan people in France, they suggest many gains can be had by working on Linux stack, which is crap right out of the box. Micah: what is the filesystem we use underneath? Alan: XFS. its stable, solid.

One of the things that is missing is space allocation.

Paul write to Arie to find out about SciDAC proposal

==== Feel out about participation in Miron/Ian proposal, which looks like it will get funded? Paul talk to Arie? Offer to interface, if nothing else... they were looking for application communities ====

LoDN

working on installing it at ORNL to work with Plasma/Fusion guys.

UltraLight and SC06

one of the problems is that Tampa is not very well connected. may only be 10 10Gig connections for whole conference, as opposed to UltraLight having 17 last time.

Root application work

Dan is making progress, but may need to talk to Micah or someone he designates about the I/O streaming aspects of IBP.