Jens, Ewan, chris, rob, david, sam, matt, brian, john, elena, Alessandra. Asynchronous SRM: seems slow, for a disk only system. Less need to pick the transfer protocol. VOMS and single namespace support. Tape support. Case for standard transfer protocols. NFS(v4) seems to be increasingly supported. If you already have Lustre, your SURL is also your TURL. A T3 may still need GridFTP for WAN transfer. How do we support the T3s, are the T3s all piggybacking on the T2s? T2s are more centrally controlled. T3s generally support only local users. Each experiment has a different definition of a T3, or a T2 supporting a different set of users. US T3s have dedicated resources. T3s (in the UK) provide a UI. T3 storage manager generally talks to the local T2 storage manager. In Manchester, the local T3 are from ATLAS and LHCb, Alessandra suggests they use CVMFS. Local groups are very different, so the advice to them will have to be T3 specific. These people normally come to hepsysman. Are there other large sites we should be talking to? Brian suggested Sussex, Warwick, Soton - other HEP sites. What about NGI/NGS type integration - this is probably a much bigger problem. Jens to advance the question? What about the eduServ cloud - we're not very cloudy but they might be interested in talking. Wild and wacky ideas - the wider picture in scientific data storage. -> for both GridPP and hepsysman. Many people will be getting big storage but will not be having grid interfaces to them. Ewan - join the dots for a national grid infrastructure, good for everybody. For the core operations groups: Three random storage documents - DPM, dCache, and StoRM installation guides. There are far too many important documents. Can refer to other people's documents. Purpose is to ensure that the core documents are up to date. [10:20:36] Ewan Mac Mahon joined [10:20:39] Christopher Walker joined [10:20:44] Rob Fay joined [10:20:44] David Crooks joined [10:20:46] Sam Skipsey joined [10:20:46] Matthew Doidge joined [10:20:46] Brian Davies joined [10:20:50] John Bland joined [10:05:25] Sam Skipsey Hi everyone. We'll start the meeting in a minute or two, I just want to let more people turn up. [10:06:34] Ewan Mac Mahon Morning. [10:08:06] Ewan Mac Mahon Let's call them Tier 4s and be done with it. [10:10:29] Ewan Mac Mahon How many Simon's do we even know? [10:10:43] Ewan Mac Mahon s/\'// [10:13:00] Ewan Mac Mahon Can't we get Jens on the phoen bridge from his mobile anyway? [10:13:11] Christopher Walker Where were the topics? [10:13:41] Ewan Mac Mahon Talks, we don't need no steenking talks. [10:13:53] Ewan Mac Mahon Discussion sessions are /way/ better. [10:14:03] Christopher Walker I've registered - at least I think I have. [10:16:27] Ewan Mac Mahon Well; obvious question - is there any new news on the Lustre front, or do we basically get it now? [10:17:47] Ewan Mac Mahon Oooh. I wouldn't mind knowing a bit more about that. [10:18:04] Ewan Mac Mahon Notably, what's it called so I can poke my Dell guys about it. [10:20:53] Elena Korolkova joined [10:22:38] Alessandra Forti joined [10:22:54] Ewan Mac Mahon Somewhat seriously, would a classic SE over a cluster filesystem actually do what we need now? [10:23:22] Ewan Mac Mahon We don't need tape interfaces at Tier 2s, and we don't need the unification of multiple servers into one system. [10:26:10] John Bland space tokens? [10:26:38] Alessandra Forti yes [10:27:07] Alessandra Forti unless experiments stop using space reservation at T2 [10:28:15] John Bland even if we had standard quotas it's still nice to have a standard interface to query/modify them [10:28:38] Ewan Mac Mahon And yes, we are a CMS T3. [10:29:02] Sam Skipsey Sure. I was just seeing how far we could logically push it, John. It might be that space tokens (even though people find them clunky) can't be entirely eradicated. [10:30:12] Rob Fay left [10:30:29] Alessandra Forti manchester too [10:31:55] Matthew Doidge We seem to refering to "Local User Facilities" [10:32:17] Matthew Doidge (back to the hepssysman roots) [10:34:07] Ewan Mac Mahon Sorry - go on [10:35:04] Ewan Mac Mahon And Oxford's local cluster is similarly separate, and is expected to remain so. [10:35:13] Matthew Doidge Lancaster's frequently finding the lines blur a bit. [10:35:33] Ewan Mac Mahon There isn't really anything that qualifies as a 'GridPP Tier 3' - the local facilities know people in GridPP, but they're not GridPP. [10:35:36] John Bland liverpool have shared DPM resources and dedicated local resources. Similarly blurred. [10:36:31] Ewan Mac Mahon I think there are useful lessons that the 'local' folks can learn from the Tier 2 experiences (I know we do), [10:37:08] Ewan Mac Mahon but I'm not sure there's much we can do in a structured and T3 specific manner, it generally comes from the usual best practice spreading stuff. [10:37:43] Christopher Walker Hadoop would be an interesting discussion [10:38:14] Christopher Walker QMUL's local cluster has had cvmfs for ages [10:38:43] John Bland ditto here, and on desktops, it's useful and doesn't take much effort [10:39:45] Jens Jensen We'll have HDFS/Hadoop at hepsysman [10:40:41] Ewan Mac Mahon Sussex are sort-of an honourary T2, they're on all the lists. [10:40:46] Ewan Mac Mahon Warwick's a good point. [10:41:37] Ewan Mac Mahon Indeed. GridPP does have a 'wider VOs' responsibility. [10:41:54] Ewan Mac Mahon So it's not completely out of scope for us, it's just a bit of a stretch. [10:43:59] Ewan Mac Mahon We need to start having some UK NGI meetings or something, rather than just being PP specific all the time, but this isn't just about storage. [10:48:31] Elena Korolkova left [10:49:06] Sam Skipsey (So, I agree with all of that, my comment about being the "GridPP Storage Group" was more a note that this is what we officially do - I'd love us to have a remit to talk to more people.) [10:50:04] Matthew Doidge Roll a dice [10:52:17] Christopher Walker left [10:52:21] Christopher Walker joined [10:52:47] Matthew Doidge I'm on the Core Ops [10:54:14] Matthew Doidge Bye [10:54:14] David Crooks left [10:54:15] Alessandra Forti left [10:54:18] Brian Davies left [10:54:18] John Bland left