Present John, Govind, Stephen, Wahid, Elena, Chris W, Brian, Alessandra, Ewan, Matt 1. Next week's talk at GridPP * TEG and conclusions * DPM workshop * StoRM and dCache plans * Capacity planning, and other planning, how we are interacting with the VOs - Chris also has a talk, on "other VOs" - - EGI suggestion for a model for VOs (Chris) - DIRAC may be interesting - catalogue synchronisation 2. Next quarter FUSE mount for LFC Data lost/corrupted - checksums - analysis into how we lose data. Study of data recovery. Losing a filesystem is bad, one Non-PP people - what are they doing? Turn into talk/report. New DPM front-ends: WebDAV, NFS4. Hadoop style processing for physics analysis. Root files cannot be split, but 3. KeyDocs are core documents for GridPP: not necessarily complete, but at least reasonably up to date, to summarise the current state. Everyone can/should volunteer to write something if they think there is gap. 4. Ricardo has been invited (via EVO) to present at hepsysman. Simon will be talking about Hadoop. Jens will put together an agenda - which is currently only an outline - and look for more "volunteers" 5. We'll meet in person next week at GridPP! Hooray! [09:58:58] John Bland joined [09:59:09] Govind Songara joined [09:59:23] Stephen Jones joined [10:00:51] Wahid Bhimji joined [10:00:52] Wahid Bhimji hi [10:00:57] Elena Korolkova joined [10:01:18] Wahid Bhimji sam said he might be a bit late [10:01:41] Wahid Bhimji It s called "Storage Directions and Issues" [10:02:00] Jens Jensen www.gridpp.ac.uk/gridpp28/programme.html [10:02:22] Christopher Walker joined [10:06:21] Brian Davies joined [10:07:28] Alessandra Forti joined [10:13:04] Ewan Mac Mahon joined [10:13:21] Ewan Mac Mahon Morning. [10:13:55] Matthew Doidge joined [10:18:46] Christopher Walker I've done my bit - all QMUL's storms are on 1.8.2 [10:21:26] Ewan Mac Mahon One of the things I think we should be doing (the whole of GridPP, not just storage) is getting a better idea what non-PP people are doing, [10:21:38] Ewan Mac Mahon and in particular the sort of people that might have been NGS sites in the past. [10:21:57] Ewan Mac Mahon Don't know how/whether we can turn that sort of thing into a metric though. [10:22:45] Christopher Walker Ah yes, cloud storage - amongst others [10:22:52] Ewan Mac Mahon And iRODS. [10:24:33] Christopher Walker https://savannah.cern.ch/task/?21779 : lfc webdav [10:26:27] Ewan Mac Mahon I think the trouble with Hadoop is that the analysis tools need to be Hadoop aware to get the real benefit. [10:26:40] Ewan Mac Mahon And that's not something we can just do on our own. [10:27:21] Wahid Bhimji I think we can - convert a final level physics analysis [10:27:34] Wahid Bhimji not talking about the heavy atlas framework stuff [10:31:32] Christopher Walker http://wiki.lustre.org/index.php/Running_Hadoop_with_Lustre [10:34:04] Ewan Mac Mahon How about the network tuning stuff? [10:34:19] Ewan Mac Mahon So the fasterdata guides?> [10:35:33] Wahid Bhimji maybe we make that more general - ie storage / network tuning [10:35:55] Wahid Bhimji with links to mysql tuning ; filesystem tuning ; network etc. [10:36:58] Ewan Mac Mahon At that point one of our 'key docs' is going to have to be a list of documentation though [10:40:22] Ewan Mac Mahon I think one of the best things we could do with for new users is big friendly letters that say "Don't Panic", followed by a suggestion to get in touch in person. [10:40:41] Ewan Mac Mahon It's very difficult to cover anything this complicated in docs alone. [10:40:43] Wahid Bhimji indeed - they can ask any kind of question they won't on the list [10:40:48] Wahid Bhimji I mean they want [10:40:57] Wahid Bhimji yes see you there [10:41:12] Stephen Jones Ewan: yes. A LOT of judgement is needed. This is not simple [10:41:15] Alessandra Forti left [10:41:15] John Bland left [10:41:18] Govind Songara left [10:41:18] Wahid Bhimji left [10:41:19] Elena Korolkova left