Attending: Chris, Ewan, Brian, John H, Elena, Gareth, John B, Steve, Mark, Sam, Dave, Jeremy, Jens Apologies: Wahid, Matt So what's good? DPM is (pretty much) sorted, roadmap is understood, upgrades are happening. It's been a quiet couple of weeks, judging from the activity (or lack of same) on the mailing list. "Big Data" for LHC: we're doing good stuff, so perhaps the good stuff should be shared more widely, eg work we can do with non-LHC as well (see chat window for Chris' report from QMUL). 400 Higgs events are a very small needle in a very big haystack. Our big data is perhaps more structured (in the data layout) and regimented (in transfers and placement) than the general concept of "big data" allows? Also Dave and Georgina (the datasets, not related in any way, shape, or form, to any real person) have been good, so maybe a catch up blog post would be useful (hint). Analysis files may go away, in, say, five years, but raw data will continue to be used. Regarding keeping track of the good stuff, FTS keeps data for at least three months but we may have records going back longer, eg UI. What's bad - other than power loss at T1... nothing much, really. Services are not fully resilient to cope with the (temporary) loss of the T1, as FTS is hosted by the T1. The data is of course hosted by other T1s as well, depending on the experiment data models. Papers for CHEP next year discussed - Glasgow leading one on "small" VOs, which perhaps QMUL (T2K) and RAL (MICE) could contribute to. Staying with Glasgow, but leaving data storage/mgmt, Dave Crooks doing one on cluster infrastructure, and Gareth doing one on CPU design. There's also the EGI forum whose deadline has been (hooray) extended to Jan. Jeremy had proposed a workshop on storage, Jens on fed id. The meeting ended with Jens thanking everyone for their work in 2012 and wishing everybody a Merry Christmas and Happy New Year; and there will not be a meeting on Wednesday 2nd, the first meeting in 2013 will be Wednesday 9th. [10:02:12] Brian Davies joined [10:02:34] Christopher Walker joined [10:03:36] Ewan Mac Mahon joined [10:04:51] Elena Korolkova joined [10:05:06] Gareth Roy joined [10:06:39] John Bland joined [10:06:59] Stephen Jones joined [10:07:48] Mark Norman joined [10:10:39] Christopher Walker http://dashb-atlas-job.cern.ch/dashboard/request.py/dailysummary [10:11:16] Sam Skipsey joined [10:12:36] Christopher Walker Dhttp://dashb-atlas-data.cern.ch/ddm2/#date.from=201112010000&date.interval=0&date.to=201212010000&dst.cloud=%28UK%29&dst.site=%28UKI-LT2-QMUL%29&grouping.dst=%28cloud,site%29&p.type=sum&tab=transfer_plots [10:13:03] Ewan Mac Mahon It's like you're on the EVO/CB radio gateway [10:14:16] Christopher Walker Data transferred. Max data transfer rates. Data processed. Events processed. are the interesting things [10:14:19] Christopher Walker has a 1.8 petabyte of storage, and runs 3500 jobs simultaneously from users worldwide. The site has analysed 24 petabytes of data in the last year, peaking at 210 terabytes in a day analysing 37.5*109 events spread over 30 million files. QMUL has processed the third most data of among the 80 Atlas Tier-2 site worldwide. [10:14:36] Christopher Walker Is what I wrote for something or other. [10:17:29] Christopher Walker CHEP papers:http://iopscience.iop.org/1742-6596/396/4/042063 [10:17:45] Christopher Walker http://iopscience.iop.org/1742-6596/396/2/022006 so far). http://iopscience.iop.org/1742-6596/396/4/042039 (Andrew McNab and Alessandra http://iopscience.iop.org/1742-6596/396/4/042054 (Sam Skipsey et al http://iopscience.iop.org/1742-6596/396/5/052075 (Wilson, Colling et al) [10:20:18] David Colling joined [10:22:10] Jeremy Coles joined [10:24:21] Ewan Mac Mahon Be interesting to get an Amazon price for 2PB of space and 24PB of transfers. [10:24:36] Ewan Mac Mahon Easy way to put a 'value' on the QMUL contribution. [10:25:01] Ewan Mac Mahon e.g. 'delivered £X worth of service for only £Y worth of actual funding. [10:25:29] Ewan Mac Mahon Shouldn't be too hard to add in the CPU on top as well. [10:27:20] Christopher Walker https://www.gridpp.ac.uk/wiki/Setting_up_a_new_Virtual_Organisation_%28VO%29 is also just a placeholder - and very incomplete, but contains some of my thoughts [10:27:27] Pete Gronbech joined [10:29:52] Gareth Roy Abstract submission for individual presentations and training sessions is extended until 11th January 2013 - http://cf2013.egi.eu/CF2013_Indico [10:30:34] Jeremy Coles Posters and demonstrations: 27 January 2013? From http://cf2013.egi.eu/call_for_participation/ [10:31:22] Jeremy Coles Presentations, training sessions: 17 December 2012