 |
Voice Conference Meeting March 1st 2000SLAC March 1, 2000 |
|
Voice Conference Meeting March 1st 2000SLAC March 1, 2000 |
Ari is working on an architecture. It is not complete yet. There is a lot of interaction between LBNL, Wisconsin and SDSC.
The client can't direct the remote staging system, however the client needs to be able to provide information to the staging system. Information is needed to be passed to provide how much space is required and how long it is likely to be needed for. It is also necessary to pass information to indicate what version of software etc. are required. It is hoped that once the code works it will be relatively easy to port to other sites.
Miron met with Globus people last week and Wisconsin will use the Globus/GSIFTP to move data around to ANL. Ed May reports that he is trying to use the fast network between ANL & LBNL. He gets between 150Mbps & 170Mbps with Iperf. It is unclear why this is so low on a 622Mbps link. Clipper is getting 71MBytes/sec between LBNL & Sandia on NTON. He is now using GSIFTP to move files from LBNL to ANL. Single transfers get 5-6MBytes/sec, 8 way parallel gets 10MBytes/sec. He is using Clipper as a cache for HPSS, and then moves the files across from Clipper to ANL (pitcairn host). The machines involved are shared but are not busy. An interesting test is to read from local disk into memory, and for this setup the performance is around 20Mbytes/sec. The setup probably needs tuning.
See http://grid.fnal.gov/ppdg/ has information on their project. Transferred about 0.5 TByte. Finished transferring all data using FTP before had a chance to get Globus installed. they were getting > 2MBytes/sec & averaged 3MBytes/sec. The bottleneck is a T3 connection. There may have been interference with Napster. They are planning to put in an HPSS system at the Indiana campus, plus plans to upgrade the T3 to OC48.
This is useful information that should be shared with, in addition maybe something on the NASA Science Grid could be presented (Bill Johnson might be a good person to presnet on the Science Grid). Given the location of the meeting in Berkeley, it shouldn't be hard to get somebody to represent the PPDG. the tentative dates are the week of April 10 or April 24.
There was a meeting following CHEP00 (see http://www.slac.stanford.edu/grp/scs/trip/italy-grid-feb00.html) where people shared information on various initiatives. There is little coordination between the countries at the moment. The big movers appear to be Italy/INFN and the UK. Grid is politically "in" in Europe and
Proposals for the EU Grid are siupposed to be in by end April 2000. There is a meeting at CERN on March 7th, 2000. There is a requirement to show that the proposals are coordinated with N. America. there may be some amplification factor if it can be shown there is coordination across the Atlantic to do with PPDG & GrYPhon.
Feedback