Attending: John B, Brian, John H, Rob, Sam, Mark, Ewan, Alessandra, Wahid, Chris W, Raul, Robert Frank, Govind, Pete Brian asked about the HDFS and Hive and Pig tutorial - we should see if we could find the materials. * Preparations for GridPP next week: * Feedback to Dave C - future issues. - Suggesting a separate storage thing. May include GPFS question. The role of standards and interoperation - SRM is a standard but not fully implemented. Not all standards are fully implemented. Reuse of code. Promotes interoperability. - "It works for us so it will work for everyone." - WebDAV will also do good stuff, but do we have antyhign which can talk to it - FTS3 can (tbc). Can also redirect... How much of it can it do what is required - see the TEG report. - Should we have a common position... - xroot is not a standard but is widely requested. - Distributed filesystem or filesystem - maybe a simpler interface - Ours is a discussion session - but we do need to discuss the SRM stuff. Could circulate prior to the meeting. - DPM support plan/doc - note that we need - Any other things we should raise - short terms stuff up to the end of the project, and future, and addressing concerns. - There is also the NGI stuff. * Coordinating feedback on the EGI GLUE schema proposal. The GLUE 2.0 schema is GFD.147 and Jens is working on feedback on a recent EGI proposal for how the schema should be published - mostly of interest to implementers of storage schemata but may be of independent interest to BDII spotters. * Federation feedback if any * What (and who) is the WLCG storage protocol working group? See also http://indico.cern.ch/conferenceDisplay.py?confId=208241 Registered people have registered interest. These are the post-TEG working groups, storage interfaces - monitor activity on sites that are moving away from SRM, spread the best practice, and come up with a list of things that the experiments really need. By feeding back into the GDB, the WG can highlight risks to the project. Who else (other than us (occasionally)) is evaluating new stuff - maybe NIKHEF/SARA. HEPiX may also be a forum for this: http://w3.hepix.org/storage/ The fact that we found the Higgs also helps us tell others that we are doing a lot of good stuff. Chris went to HPCSIG and they were also interested. Do we get value for money in the public funding? GlobusOnline is easier to use than FTS, if we could move away from SRM. There has been a request for GO to support SRM as well. AOB Did anyone do tests with Alessandro di Salvo on DPM? Problem with 32 bit libraries. Wahid thinks it's been fixed, will forward details. Sometimes this information is around but doesn't always get communicated to the right people. [11:16:08] Alessandra Forti joined [11:16:31] Wahid Bhimji joined [11:17:27] Wahid Bhimji sorry audi o issue [11:20:13] Ewan Mac Mahon Standards are lovely. There are just so many to choose from. [11:21:18] Christopher Walker joined [11:21:20] Wahid Bhimji I'm not sure the xrootd push - shows a keenness for standards [11:21:31] Wahid Bhimji left [11:21:34] Wahid Bhimji joined [11:22:02] Christopher Walker We'd be better going for http/https IMHO [11:22:33] Sam Skipsey Well, there are at least one or two people looking at that in ATLAS, Chris. [11:22:59] Wahid Bhimji no [11:23:46] Wahid Bhimji FTS3 [11:24:46] Ewan Mac Mahon So at that point you'd mount the LFC as a webdav server and an ordinary webdav client would understand the redirects to the real SEs? [11:25:08] Wahid Bhimji in theory - not demonstrated [11:25:35] Ewan Mac Mahon That sounds cool. Let's have one of those and some storm/lustre setups. [11:26:31] raul lopes joined [11:28:28] Ewan Mac Mahon I think if the lcg-utils/gfal/etc. interfaces could talk the appropriate protocols it might be possible to move the backends across without totally stuffing the users. [11:28:58] Ewan Mac Mahon In the same way that you can have file:// protocol or rfio:// access and the end user doens't need to care - just the experiment framework. [11:30:46] Robert Frank joined [11:31:03] Govind Songara joined [11:32:00] Ewan Mac Mahon Well poke Pete about it again. [11:34:22] Christopher Walker So the question is "What's the future of HEP storage - disk specifically, and how does that influence what should be done with DPM" [11:34:49] Ewan Mac Mahon I think we do need to get it down to a list of questions that need answers. [11:35:14] Ewan Mac Mahon The current slides are a bit more along the lines of some stuff we could maybe do. [11:35:52] Brian Davies sofa poilitics! [11:36:09] Wahid Bhimji sure its not a big deal I was just highlighting that the wider eyes [11:36:16] Pete Gronbech joined [11:38:03] Ewan Mac Mahon Also, it's two half hours on the agenda, isn't it? [11:38:16] Ewan Mac Mahon 12:00-12:30 and 12:30-13:00 [11:42:28] Christopher Walker Small VOs will want something easy and standard. [11:42:39] Christopher Walker Standard idoms [11:44:44] Jens Jensen http://indico.cern.ch/conferenceDisplay.py?confId=208241 [11:47:36] Christopher Walker In principle, Mark Mitchell is the GDB rep, but Jeremy has suggested that someone more interested in storage might wish to go. [11:48:37] Alessandra Forti sites don't move without the experiments saying so [11:48:53] Christopher Walker In the absence of anyone else here wanting to go, then I'll probably volunteer (but I did go this month). [11:49:50] Alessandra Forti they are the only ones seriously using space tokens [11:50:03] Alessandra Forti if they don't get rid if that they don't get rid of srm [11:50:08] Alessandra Forti at least at T2 [11:50:10] Alessandra Forti s [11:53:24] Jens Jensen http://w3.hepix.org/storage/ [11:56:13] Ewan Mac Mahon There's a degree of wilful blindness I think [11:56:39] Ewan Mac Mahon (some of) the HPC folks think that their shiny exotic kit is the best thing going and they don't like listening to us [11:57:00] Ewan Mac Mahon when we say how we can do huge things on commodity kit for very small amounts of money. [11:57:05] Ewan Mac Mahon (relatively) [11:57:31] Ewan Mac Mahon I mean - DIRAC's setup funding was something like £25million, wasn't it? [11:57:53] Ewan Mac Mahon And does it have a G in the middle, or is it an E? [12:02:14] Ewan Mac Mahon This is the EMI 2 WN client libs problem. [12:02:20] Ewan Mac Mahon Or non-problem as it may be.