GENIUS Meeting 10 Minutes

From RealityGrid

GENIUS Meeting 29/11

Item 1:

UCL: LF, PVC, SM, OK, SZ, MM RAL: AR Leeds: JL Ox: SY Manchester: RH, RP LSU: SJ, DK NCSA: DM, TC SDSC: ?

Item 2:

SM: Demos went well, big thanks. LONI demos went well using MPIg/Viz etc.

Item 3:

PVC: DEISA looking for co-reservation mechanisms. Wants to work with us on deployment of HARC. Several senior members of DEISA confirmed this. DESIA looking for funding for DEISA2. DEISA want to have a co-res demo by April.

PVC: Thomas Soddermann not able to come today, but should usually be able to come.

PVC: DEISA plans to look at HARC and see how long it will take to get in to a production phase. People from DEISA will attend the Mardi Gras conference, and hope to meet LSU people there.

DK: TS has exchanged some emails with DK.

PVC: We may see some DEISA representatives in this meeting in the future.

Item 4:

SY: Haven't been any changes - link still in place and working.

SM: Haven't got the UCL-TG (and through to LONI) link up yet. Will push the application through now.

Items 5, 6 & 7:

RP: Haven't had time to look in to MPIg problems on NGS; worried that the problems are due to MPIg/Globus bugs.

SM: Have had some problems on TG - LONI is best.

OK: For small testing runs, TG problems have not been related to MPIg, but having problems with larger runs.

SM: Want to try to analyse where bugs lie, and invoke the relevant developers.

RP: Will wait for other platforms to become more stable, before trying to get to work on NGS.

OK: Number of problems with production runs. Suspects that the problem is with cross site LAMMPS, has got to the point where LAMMPS starts up. Previously problems with nodes not knowing where libraries are- config issues with backend nodes.

SM: Will try writing some simple MPIg test cases, and use BT's test cases, to isolate what is happening.

JL: Has tried to solve some of the Globus bugs on Leeds, and has asked OK to try running some test cases. Needs someone to test changes.

AR: Will look in to OK's NGS application.

AR: Can support a VO for managing hour allocations. Will send info to PVC on how to set up a VO.

DM: Can SM keep DM in the loop with any MPIg tests on TG sites. Do have some tests the run on TG sites, want to make sure they are running the correct (automated) tests.

DM: Will take the lead on the problem of not getting output from jobs run with output. Will co-ordinate testing with OK on large scale HARC reservations not working.

SDSC: Might have made some changes to scheduler, due to SMs TG ticket. Looks like problem is with PBS's inability to copy output to the GASS cache.

JL: Had some complaints with the same kind of thing happening on PBS with HARC.

OK: Problem got work when NCSA upgraded from NFS to GPFS

SM: Thinks there is a timing issue.

Item 8:

PVC: Likely to have HARC on Lonestar soon.

SZ: Summary of AHE development

RH: Going to use platform grant money to do more work on ReG. It could be made more accessible to more people, and simplified. Hopefully will just notice things getting easier.

Item 9:

PVC: How do we track this, and also changes to systems made by sysadmins just before a demo etc. How can we identify the people who need to be told across sites.

DM: Will think about it, and work out who's attention this should be brought to.

PVC: Do we need to have themed meetings (e.g. MPIg)

Item 10:

PVC: Would like to continue with 2-weekly meetings. Could do half hour general meeting, and half hour specific meeting, or alternatively/additionally run satellite meetings.

RH: Full meetings do help keep people involved in project. Themed meetings useful as we get deeper in to project.

RH: Keep general meetings every two weeks.

RM: Themed meetings could happen more spontaneously, with full meetings happening regularly.

PVC: Could set up sub email lists.

AR: Need to have a workgroup leader, and main meetings at least once per month.

PVC: Could set up sub-theme mailing lists, and will identify who they should be.

RM: Forums could be a good solution to this.

RP: Thinks for forums would be a good idea

AR: More than welcome to use NGS forum as a tool

SJ: Thinks current single mailing list is good and effective.

SY: Will we put together a test harness to automate the cross site tests.

SM: Yes, but not sure how.

AR: Inca is a testing framework used to monitor sites; basic framework which can have arbitrary tests run from it.

DM: TG uses Inca too.

SM: Will look in to it.

PVC: Will stick with one mailing list, and possibly multiple fora.

DK: DK and SM will be at e-Science meeting in India for date of next meeting.