Attending: Brian, Chris, Elena, Govind, Jens (C+M), John B, John H, Matt D, Winnie, Sam, Steve, Tom, Wahid, George V, Ewan Special guest star: Marcel Stanitzki from DESY representing ILC Also starring: Rob Appleyard from RAL 0. Operational blog posts No operational issues. Reminder to people to blog. We will discuss storage outside the fw next week. 2. Marcel Stanitzki from DESY will join us and report on the ILC data model and answer questions about all things ILC: Marcel gave a presentation (coauthored with Jan). ILC to do linear acceleration of e+,e-, as leptons, unlike hadrons, have quite small wheels and don't do corners very well. Single event per train, so different from LHC. There is a TDR, see link in chat. Mainly production (and reconstruction) at T1, expecting user data at T2s. Using ILC Dirac which is an offshoot from LHCb's Dirac. Japan have expressed an interest in hosting the ILC, in the Kitakami mountains - which should be earthquake safe, being a solid block of granite. Physics by 2030. Estimating 9PB/yr of raw data, which should be easily manageable by then. A question is whether to have another T0 or just use T1s. Generally expecting same model as Belle2, where they send a complete copy of data to BNL. From Japan. It is possible a cross-pacific cable will be laid between US and JP; Brian says there will be a northwest passage one but mainly of interest to finance. Marcel also thanked GridPP for the contributions. Will data be archived or will they be working repositories? Probably one online copy is needed; the rest can be nearline, but hard to predict technology advances by 2030. (LEP was quite User output expected to use ILC catalogue. 1. Report from xroot workshop - Rob Appleyard from T1 will join us and tell us tales from the xroot workshop. People extol their FAX, AAA, etc. There was a discussion whether storage allocations could be configured centrally instead of individually by the sites - several people raised this point [perhaps not coincidentally, the EGI FedCloud has just asked the same question - JJ] There was a xrootdfs, via FUSE, with a flexible security model. There was a file caching proxy (as in web proxy, not certificate proxy), available for Xrootd4 only. There was a discussion of remote debugging: if you can enable remote debugging in a controlled way, it would greatly speed up the debugging process - instead of having email pinging back and forth, a remote debugger can access logs and cores, in a sufficiently restricted way (so not just giving root access.) CERN wrote a new xroot4 client; some of the client tools backward compatible. Shiny. It would be good to have an improvement of logs, because logs are too odd for RAL to use ElasticSearch. Sebastien Ponce had written an xroot-to-CEPH interface, which, unlike many of the other interfaces we have seen, allows the data to be read back also through other interfaces. Idea is to talk to the RadosStriper which is in Giant. Sebastien also has a GridFTP plugin for CEPH. Also for EOS, but this is of less interest to GridPP - here, every option can be passed in the URI!. Sam is writing the equivalent xroot plugin for GFAL2. Not just HEP sites using xroot, also Qualys and NERSC. http://indico.cern.ch/event/330212/session/6/contribution/27/material/slides/0.pdf 3. For today's (ie Wednesday's) GDB, there is a bit about "DataCloud European Project" which is scheduled to start just after this meeting: http://indico.cern.ch/event/319744/ 4. Does this CFP look interesting: http://press3.mcs.anl.gov/iasds15/ 5. AOB NOB Chat log Tom Whyntie: (11/02/2015 10:01:14) Mornin' Matt Doidge: (10:01 AM) Good Moaning. Govind: (10:02 AM) Yes Matt Doidge: (10:02 AM) I've always run our DPM with iptables only. Govind: (10:03 AM) We can't afford expensive firewall with 10GB link so we have to move it now.... so discussion will be useful. Chris Brew: (10:06 AM) Someone came in and asked me a question just I was about to speak up on firewalls. Here's some through and bypass numbers for the RAL firewall Firewall Bypass No. Streams Out In Out In 1 96.0 Mbits/sec 70.7 Mbits/sec 316 Mbits/sec 103 Mbits/sec 184 Mbits/sec 65.5 Mbits/sec 225 Mbits/sec 484 Mbits/sec 405 Mbits/sec 359 Mbits/sec 435 Mbits/sec 589 Mbits/sec 3.32 Gbits/sec 5.53 Gbits/sec 2 310 Mbits/sec 281 Mbits/sec 5.92 Gbits/sec 7.16 Gbits/sec 5 551 Mbits/sec 1.01 Gbits/sec 6.59 Gbits/sec 7.14 Gbits/sec 10 1.07 Gbits/sec 1.41 Gbits/sec 7.09 Gbits/sec DNF Brian Davies @RAL-LCG2: (10:09 AM) slidews now uplodaed to https://indico.cern.ch/event/313426/material/slides/0.pdf Jens Jensen: (10:10 AM) Thanks, Brian. Ewan Mac Mahon: (10:25 AM) Do we think tape will still make sense that far in the future? 'Archive' type disks are really impressively cheap now. Jens Jensen: (10:28 AM) Climate is looking at tape ahead to 2030. The main challenge is bandwidth to tape, not storage capacity Ewan Mac Mahon: (10:29 AM) Indeed; it's a detail of implementation, not a feature of the model. But I think we are getting to the point where we need to be careful about the terminology. Samuel Cadellin Skipsey: (10:30 AM) Oh, I agree - thanks for reminding me to be careful. I do keep saying "tape" when I mean "archival" Ewan Mac Mahon: (10:30 AM) Someone offered me a box with ~0.5PB for ~20k GBP the other day; things are shifting again with disks. John Bland: (10:32 AM) ewan: was that just a carboard box filled with disks? Ewan Mac Mahon: (10:32 AM) This was the new Seagate shingled 'archive' disks - slow but speactacularly cheap. :-) It was a 36bay Supermicro with a 36bay JBOD box, two LSI RAID cards and a 40Gbit (!) network interface. Samuel Cadellin Skipsey: (10:33 AM) Well, with that capacity, you need the 40Gbit interface :D Ewan Mac Mahon: (10:34 AM) So it was 72*8TB = 576TB raw. I'm not quite sure what I'd do with one, but it was a really interesting looking thing, The price on these drives is a fraction of the price of 'normal' high capacity disks. They really do seem to be a step change in that line that's been very slowly bringing bulk disk towards tape prices. Marcel Stanitzki: (10:36 AM) http://www.linearcollider.org/ILC/Publications/Technical-Design-Report Rob: (10:37 AM) I'm here! Awkward. Slides from the workshop are located here if you need more information on a given subject: https://indico.cern.ch/event/330212/contributions Ewan Mac Mahon: (10:40 AM) Ah, and I've just checked the price on said box and it's more like £28,000 once you put the VAT in, but still. wahid: (10:50 AM) bye - thanks - cu .