Minutes of storage EVO meeting 17 Feb 2010 Present: Glasgow: Sam Liverpool: Stephen, John Manchester: Alessandra Edinburgh: Wahid Lancaster: Matt QMUL: Chris Imperial: Duncan Bristol: Winnie Sheffield: Elena (from 10:17) RHUL: Govind RAL T1: James, Brian, Jens (chair+mins) 0. Action review, see below. 1. Matt to talk us through his 10 Gig (not-E) experiences: Slides at http://www.hep.lancs.ac.uk/~msd/LancsNetUpgrade.pdf Additional questions: size of pool in relation to network: high perf network also gives lower latency. Chris had iperf maxed out at 5 Gb. 10 GigE cards seem expensive by themselves but when viewed as a fraction of systems cost seem more reasonable. Thus it may be better to get them for new systems rather than retrofit them into old ones. 2. Review of checksumming planning See http://www.gridpp.ac.uk/wiki/File_Integrity_Testing 2+3 together: many tasks keep popping up again and again, and often an action is opened which stays open for years while some poor person hacks away at it (or it could be a slow person of course :-) Often it is because the problem is more difficult than originally expected. It is better to treat it like a difficult problem and plan how to attack it. The file integrity testing is a good example (see wiki). 4. AOB Jens will be out next Wed. NON-LOW ACTIONS 338 04/11/2009 Test and compare performance of DPM vs StoRM/GPFS Sam+Wahid Med Open Ongoing Maybe reword to cover "file access" 348 02/12/2009 Circulate 10GigE stuff Matt Med Open "soon" Done, see agenda item. 356 13/01/2010 Check remote DPNS access to sites Brian High Open Ongoing Done. 359 10/02/2010 Kickstart mini plan for checksumming Jens Med Open Done, see http://www.gridpp.ac.uk/wiki/File_Integrity_Testing 360 10/02/2010 Create files for Brian to close dpns testing Sam High Open Done. 361 10/02/2010 Document existing tests in wiki Jens Med Open Not done yet. 09:58:13] John Bland joined [09:59:44] Stephen Jones joined [10:00:03] Wahid Bhimji joined [10:00:36] Queen Mary, U London London, U.K. joined [10:01:42] James Thorne joined [10:01:51] Matthew Doidge joined [10:02:57] Govind Songara joined [10:03:19] Matthew Doidge slides for my presentation: http://www.hep.lancs.ac.uk/~msd/LancsNetUpgrade.pdf [10:04:13] Brian Davies joined [10:04:21] Brian Davies apologies for tardiness, evo being evil [10:06:31] Wahid Bhimji what about Role [10:09:47] Alessandra Forti joined [10:13:46] Jens Jensen THere is a lot of feedback noise [10:14:16] Wahid Bhimji can't hear at all - perhaps reconnect [10:14:21] Alessandra Forti me too [10:14:28] Alessandra Forti or me niether [10:14:33] Brian Davies matt , we can not hear you [10:14:34] Jens Jensen Hi Matt, are you still there? [10:16:25] Alessandra Forti left [10:16:31] Alessandra Forti joined [10:16:34] Brian Davies noise cancellation? [10:17:18] Elena Korolkova joined [10:18:07] Alessandra Forti left [10:18:48] Duncan Rand you could try skyping into the phone bridge [10:21:32] Alessandra Forti joined [10:23:40] Alessandra Forti left [10:25:00] Alessandra Forti joined [10:28:13] Matthew Doidge my apolgies for the dodgey audio [10:28:24] Wahid Bhimji no probs - interesting talk [10:28:54] Jens Jensen http://www.gridpp.ac.uk/wiki/File_Integrity_Testing [10:30:20] Queen Mary, U London London, U.K. At the GDB, one of the T1s said they had had a problem with a RAID card - and initial checksums on arrival were fine - the data was in RAM. You need to read the data from disk... [10:31:14] Sam Skipsey Chris - indeed, this is the difference between "deep" checksumming and checksumming in our nomenclature. [10:34:20] Duncan Rand http://panda.cern.ch:25980/server/pandamon/query?mode=dsPopularity [10:34:59] Duncan Rand click on the column header to sort by that header [10:35:28] James Thorne left [10:35:36] Wahid Bhimji thanks - bye [10:35:40] Winnie Lacesso left [10:35:41] Govind Songara bye [10:35:42] John Bland left [10:35:42] Elena Korolkova left [10:35:42] Matthew Doidge left [10:35:44] Govind Songara left [10:36:07] Duncan Rand left