GENIUS Meeting 15 Minutes
From RealityGrid
GENIUS meeting 28/2/2008
Item 1:
EPCC: SB
Man: JB, RH, JM
NCSA: DM
PSC: SB
SDSC: KY, Martin
UCL: PVC, SM, SZ, OK, GD
LSU: DK, SJ, PA
Item 2:
PCV: Thanks and best wishes to JM.
Item 3:
DSK: Talking to DESIA, and working to get HARC deployed. Thinks it will be on an IBM P6:
Action item: Get in touch with Thomas Soddermann re getting access to machine with HARC deployed.
DSK: Will be deployed at one DEISA site initially.
Item 4:
SM: SM, MM, and GD visiting NHNN every Tuesday. New angiography suite will be up and running by 10/3. Have met with UCL and NHS networking staff to get necessary network connections in place, and also getting hold of patient data. Have contacted Siemens re integrating GENIUS software with their software.
PVC: Moving forward at a reasonable pace there.
Item 5:
SB: Expect EPSRC to extend life of HPCx by 14 months. Service will be refocused to do things that aren't possible on HECToR. HARC already deployed on HPCx - seems to be working. Current reservation config: 6 hours duration, 128 CPUs, 48hrs notice. Get in touch with SB if this doesn't meet needs.
SB: Keen to support back end nodes. Extension will prob be refreshed to Power 6. This would be a good time to reconfigure nodes.
SB would like to do it earlier. Problem with allocation of IP addresses - not enough. Has spoken to David Salmon about this - could be supplied through Janet/DEISA.
PVC: Have kept large part of allocation on HPCx rather than transfer to HECToR.
PCV: Would be interested in working with HPCx to implement emergency computing infrastructure.
SB: Send details of how this is done on TG.
SB: Never been configured for checkpointing jobs, so not sure how jobs would be pre-empted.
KY: Loadleveler has ability to checkpoint jobs and keep them on the nodes.
PCV: Want to deploy HARC on Lonestar and Ranger. Don't think Warren at TACC has started deploying yet.
JM: Doesn't think reservations are enabled yet.
DM: Concurs. Obstacle at TACC is the implementation of advanced res.
KY: Trying to get in on Ranger, but lots of work on getting Ranger in to production.
PVC: Difficult to get urgent computing job with > 1024 cores. Getting such on Ranger doesn't sound like a difficult request, so keen to do it. Working with Beckman et al to move this forward.
SM: Problem with threaded Globus & MPIg on IBM frames.
PA: Test getting multithreaded Globus working on Power 5 frames.
SM: Asked for details re LONI network. Would like to give details on how lightpaths operate.
DSK: CC him on email and will assist.
Item 6:
SM: Opening up back ends on Cray machines is difficult due to interconnect and number of cores. Would like to look at getting PDIO for steering, and using RSIP stuff.
SBrown: getting it to work with steering software should be straightforward. Will talk to developer of PDIO on getting it to work with HECToR.
SBooth: Have Globus etc on test/dev frame on HECToR, and should be able to get RSIP working on there. System available to support project. Will send SM documentation on RSIP.
Item 7:
MM: Performance of new ray tracer is 3 times state of the art?
MM: From next week will benchmark pulsatile based simulations with GD. Updated version will be in CVS.
RH: Not much to report re ReG update. Work is ongoing. Nothing specific is finished.
SZ: AHE client updated to use latest HARC API.
Item 8:
PVC: Invited to contribute to special edition of CTWatch. Hopefully our article on medical urgent computing should hopefully be available on web soon. Preparing to submit to TG conference. RH is heading an article for Capability Computing.
SB: No plans for similar HECToR machine.
RH: Will liaise with Andy Porter to plan high-level article for CC. Use material we already have, but pitched at HPCx. Article deadline end March.
MM: With new benchmark results, plan two new papers in next month or so.
PVC: AHM - PVC chair of meeting this year, and location changed to Edinburgh. Expects lots of participation from GENIUS meeting participants. Hope people here will help with distribution of calls etc. Have secured 2 issues of Phil Trans to feature papers.
PVC: SC08 - in Austin in Nov.
JM: Where do we plan to base demos at SC08? Doesn't think there will be a UK stand this year.
SB: Thinks EPCC will have a stand.
SBrown: Co-chair of science track for TeraGrid. Encourage people in the meeting to contribute. All work done can be submitted to other journals to publish - TG won't hold copyright. Anyone who wants to be a reviewer please volunteer.
Item 9: AOB.
None.
Item 10:
Next meeting 27/3/08.
PVC: SM visiting Riken and Naregi in Japan next week.
DSK: Working with Naregi an Riken.
Item 11:
PCVC: subsequent meetings will be every other week.
