Attending: Matt RB Duncan John B John H Sam Steve Wahid Brian Gang Raja Elena Ewan Tom Govind Chris W Raul 0. 100% of target achieved in previous quarter (probably) - all green - Pete will be so proud of us he will buy us all pizza and beer. Speaking of 100%, Ewan has done 100% of interesting work in this quarter (so far) - and no April fools. 1. Operational problems Brian reports problem with Cambridge seeing slow transfers. Cambridge had doubled connection to JANET, so would expect better performance, although this is recently, whereas the problem Brian reported dates back a bit further. TCP tuning might be the way forward (see links in chat) - there was a bug for multiple streams transfers, and it was recommended that sites set their default buffer window higher by default. Oxford also found the need to do stuff, and had problems in particular with transfers to BNL. Brian will investigate and follow up. For now, there is no need to follow up with anyone outside the UK. Glasgow moving from cfengine to puppet. 2. GridPP summary http://gridpp-storage.blogspot.co.uk/2014/03/storage-thoughts-from-gridpp32.html In general big (LHC) experiments are happy; smaller experiments (non-LHC) sometimes confused. Also, with LFC being "de-emphasised", there is a need to manage the implications. How are non-LHC VOs supported: need not just effort but also understanding; maybe GridPP should dedicate some effort to this, and not just Tom's and Chris' spare time... however, responsibility for the integration should remain with the VO, as they understand their proposed work better than we can (in general). Large VOs are changing their data models. Will impact deletion rate, dark data discovery, and suchlike. 3. Jens gave a brief summary of ISGC which had a good GridPP presence. http://gridpp-storage.blogspot.co.uk/2014/03/highlights-of-isgc-2014.html There were tutorials on DIRAC and dCache, both of which seem easy enough to use once you get your security credentials sorted (dCache can run without security....) and of course in a tutorial there is always a tradeoff between doing stuff and getting things set up. Interesting how LHC work binds countries together, with all the other things they need to do: set up CAs and join IGTF, join operations support, etc. CEPH also mentioned a few times - RAL and FNAL proposed to collaborate, but this should be an open collaboration where others can join (Glasgow?) See also blog post and chat. 4. AOB Meeting continued with a technical xroot discussion led by Wahid. [10:00:18] Matt Raso-Barnett ah i was [10:00:26] Matt Raso-Barnett must test my mic [10:06:23] Wahid Bhimji perhaps worh checking the tcp tuning parameters in case [10:06:30] Wahid Bhimji they reverted [10:07:02] Wahid Bhimji https://www.gridpp.ac.uk/wiki/UKTcpTuning [10:17:25] Jens Jensen http://gridpp-storage.blogspot.co.uk/2014/03/storage-thoughts-from-gridpp32.html [10:17:39] Jens Jensen https://indico.cern.ch/event/299622/ [10:21:04] Wahid Bhimji dirac data management [10:21:18] Sam Skipsey ...wait, another DDM? [10:21:31] Raja Nandakumar Actually called RMS - the replica management system [10:22:20] Jens Jensen Another TLA [10:23:32] Ewan Mac Mahon That's not 'slightly dodgy'. [10:23:59] Raja Nandakumar Really? [10:24:14] Raja Nandakumar [10:24:38] Tom Whyntie Me too - having mic problems [10:27:11] Ewan Mac Mahon This is, BTW, the sort of thing that could have been an interesting and productive discussion at the meeting if the meeting hadn't been such a disaster for having discussions. [10:27:30] Sam Skipsey quite, Ewan [10:28:13] Sam Skipsey (Well, not quite: Dan still has stuff that goes beyond Dirac for some specific things he cares about.) [10:30:32] Raja Nandakumar Dirac will be happy to assist in the beyond things I suppose [10:30:41] Jens Jensen http://gridpp-storage.blogspot.co.uk/2014/03/highlights-of-isgc-2014.html [10:30:41] Raja Nandakumar Apologies - got to go now. [10:30:58] Wahid Bhimji Brian - are you using FTS3 for ATLAS at the moment [10:31:08] Wahid Bhimji what was the status there [10:31:51] Ewan Mac Mahon EGI Federated Cloud - not really EGI, not a Cloud, and not Federated. [10:32:00] Ewan Mac Mahon Other than that, it's awesome. [10:32:32] Sam Skipsey I missed the Webinar on it (The EGI Federated Cloud) this morning. [10:32:52] Brian Davies FTS3@RAL is being used by ATLAS for UK cloud only.. ATLAS are using FTS3@CERN for other clouds. plan is to move more of the other clouds back to usding FTS3@RAL soon(ish) [10:33:09] Wahid Bhimji OK can you try xrootd transfer [10:33:11] Ewan Mac Mahon Do we want to migrate away from the e-Science CA? [10:33:14] Sam Skipsey I am surprised that the EGI Federated Cloud claims elasticity - I would have thought they'd have capacity issues. [10:33:15] Wahid Bhimji with FTS3 for ECDF [10:33:24] Sam Skipsey Ewan: I think the question is one of resources? [10:34:15] Wahid Bhimji @Govind - re xrootd - so local transfers are OK now for atlas [10:35:04] Sam Skipsey I dunno, "Interware" sounds like an item of clothing. [10:35:49] Tom Whyntie Yes [10:37:26] Sam Skipsey Those of us who know how to pronounce Greek words [10:37:33] Jens Jensen [10:37:58] Ewan Mac Mahon I'm pretty sure everyone that calls it keph rather than seph is in this meeting. [10:38:05] Sam Skipsey This is true. [10:38:06] Tom Whyntie Great, thanks