Attending: John B, Brian, Jens (chair+mins), Winnie, Steve, Marcus, MattD, Sam, Tom, David, Raul, ELena, John H 0. Operational blog posts httpd - still not resolved. Maybe sit down next week and look at it together, if it is not resolved by then. No blog posts since April 20 (and Brian and Jens are not allowed to blog till Friday next week) 1. ATLAS stuff - what happened to that hammercloud test and all that other stuff? Cache testing ongoing; need feedback from Oliver. Alastair still needs to report on T2C testing but is also obviously busy with other priorities; could anyone else do it? 2. Future T2s - what can we conclude so far? Can we write anything up yet? Is a future T2 running DPM or StoRM or dCache? Does it use SRM? Do all T2s have their own storage? Is storage at T2Cs cache for some fuzzy level of cachiness? CVMFS to distribute data (cf. GDB)? DIRAC? Object stores as local cache? Now is a good opportunity to write up some stuff about where we see ourselves five years from now, even if not everything is certain. At least we can highlight what the uncertainties are. Steve points out we will have manpower restrictions so future "solutions" will need to be operated with relatively little effort (which in fact was the original use case for DPM, and probably also StoRM :-) and Sam points out we will also have resource restrictions in general, so will need to do more with less. Some uncertainty in the experiments feedback, ATLAS and CMS have different requirements; LHCb is moving to heavier T2s from the lightweight ones used so far. T2s may come in flavours ranging from all CPU to all storage and with caching T2s in the middle. Also ATLAS's model seems to be evolving with nucleus and sattelite sites. The famous Lisbon meeting essentially ended without agreement. There are also the non-LCG experiments; good feedback to UKT0 from DiRAC and LSST; there is an opportunity for us to tell UKT0 and others that we can do data infrastructure (cf the Science DMZ discussions) - once across the barrier, GridPP works well - and also conversely we look at technologies in storage and networking which is relevant for our future evolution. How are priorities to be weighted? There will be political constraints, and technical. Steve sent some thoughts to Sam; Sam will write something and send to a small number initially; Jens will start a Google doc or something. Also in the big-picture-y department, how will we link (if at all) to EGI, EUDAT, PRACE? Will be support Globus? What about Open Science Cloud? What role of buzzwordy things like networking technologies, public CSPs, BYOD, and there is the future of networking; for example, in the US it is sometimes possible to allocate bandwidth and when it is guaranteed, one need not run TCP. There is the DMZ stuff as well. 3. Loose ends: hepsysman is next week - lots of site reports, ZFS report https://indico.cern.ch/event/518392/ Opportunity to meet - anything we want to hear about in the ZFS talk, please let Marcus know. Licensing is probably the main concern. (Hence btrfs but it has not been progressing as rapidly as one could have wanted.) MSDC update Aforementioned good feedback from LSST and DiRAC; need to make the case for GridPP-as-a-data-infrastructure (hence the above doc). Funding for AAI "pathfinder" project. 4. AOB Tom Whyntie: (15/06/2016 10:36:43) LSST regarded as a huge success for GridPP, by the way - some great quotes from the LSST PM team Tom Whyntie: (15/06/2016 10:36:43) LSST regarded as a huge success for GridPP, by the way - some great quotes from the LSST PM team Sounds good - thanks, see you next week!