Present: Chris, Duncan, Elena, Ewan, Gang, John B, John H, Matt, Raja, Rob, Steve, Gareth, Brian, Sam, Raul, Jens, and Matt V from RALT1 Apologies: Wahid 1. GridPP 5 proposal Wahid had written up some stuff about the future of GridPP in WLCG - the growth of the need for disk, etc. Jens had some notes about the roles of the group. What is the role of GridPP (SaDM) in the future? Clearly we are primarily supporting LHC, but other projects are looking to us for managing "big data" and clearly we should keep talking to/with them. In bioinformatics, for example, there are people writing data to tape and fedexing them (as opposed to PhEDExing them). In terms of user communities there are also values added by having other communities use our infrastructure, provided they could use the things as they are and only need some handholding, as opposed to customisations. Non-HEP communities seem to have a higher resistance to the learning curve associated with using certificates, command line tools, 2. Round table discussion - storagey things Postponed to next week, or maybe later... 3. Technical Topics Could do iRODS, Globus Online, ... 4. Digital preservation in HEP - Matt gave a quick presentation of recent work at CERN (or led by CERN) on preservation of files. Work focuses on the preservation at the bit level, and ignores the format migration and management, but may at least attempt to preserve tools and binaries in addition to the files. [10:05:51] Steve Jones We used to say "Never underestimate the bandwidth of a car boot full of tapes." [10:08:54] Tom Whyntie So is the question here: "do we put resources in the GridPP5 bid for dedicated (G)UI development?" [10:11:59] Sam Skipsey Apologies for my tardiness. [10:12:07] Ewan Mac Mahon We do put some effort into inviting people to come knocking on the door, we're not just sitting around waiting. [10:12:35] Ewan Mac Mahon I think it's more a matter that we're not going to put effort into something without at least a bit of an indication that someone's actually interested. [10:12:35] Steve Jones The nice thing about standards is that you have so many to choose from. [10:12:47] Steve Jones Another Tannenbaum quote. [10:14:19] Sam Skipsey Parrot, right? [10:17:10] Ewan Mac Mahon This is just another case of the general question of whether GridPP is aiming to by a generic UK HTC resource, [10:17:16] Ewan Mac Mahon or the UK's contribution to WLCG. [10:17:49] Ewan Mac Mahon We can do either, but someone needs to decide which it's going to be. [10:18:29] Steve Jones Yes, Ewan. There are trade-offs to be made. [10:20:32] Ewan Mac Mahon Our resources are actually a lot more 'normal' in many ways than the environment on most HPC clusters, which differ more from each other, and from people's desktops. [10:20:49] Ewan Mac Mahon What you see on an SL6 WN is a very vanilla Linux environment. [10:21:19] Ewan Mac Mahon With a few knobs on. [10:23:18] Steve Jones Yes - there is no such thing as a free lunch (not Tannenbaum). [10:24:17] Sam Skipsey I never thought I'd see an old socialist like you quote Heinlein, Steve [10:25:17] Steve Jones [10:26:20] Ewan Mac Mahon We have about as high a degree of assurance and tracability than anything else. [10:26:34] Ewan Mac Mahon Er, that changed emphasis half way through. [10:30:09] Steve Jones Sadly, we had to ban biomed because they launched 8,000 jobs on us at once, repeatedly. I wish they'd throttle that feeback problem. [10:30:24] Sam Skipsey Aye, we had a similar thing happen. [10:31:11] Ewan Mac Mahon On user requirements - it depends what you mean by 'users' - a person will always want security to be unobtrusive to the point of non-existence. [10:31:32] Ewan Mac Mahon But their management may want data to be secure and accesses logged. [10:31:34] Sam Skipsey Indeed: hence why everyone still uses passwords on the internet. [10:32:46] Ewan Mac Mahon Except for the (minor) fact that they're starting to not - where accounts are valuable and the risks of losing control of them significant enough, people are somewhat demanding better. [10:32:59] Ewan Mac Mahon e.g. two factor auth for Google, twitter & github accounts. [10:33:18] Ewan Mac Mahon (And facebook?) [10:33:25] Sam Skipsey Indeed. But that took a long time to happen. (Of course, I turned on two factor for everything that supports it) [10:33:32] Sam Skipsey (Yes, and Facebook) [10:33:55] Sam Skipsey also: I am not convinced that buyin to two factor is actually significant. [10:35:16] Ewan Mac Mahon It did, but I'm not sure it's technical time so much as cultural. There used to be a view that 'the internet' was a separate and less real world, and essentially inconsequential. As people come to value their online existences more, they're prepared to wear more security. [10:39:16] Ewan Mac Mahon That's striking. The current view seems to have been moving much more strongly to active curation for long term preservation and away from the traditional 'lock tapes ina safe' approach. [10:40:11] Ewan Mac Mahon I'm deeply sceptical that keeping tapes for 30 years is a sane approach - you might have tapes in perfect condition, but I wouldn't rate your chances of finding anything that can read them. [10:42:37] Sam Skipsey http://indico.cern.ch/conferenceOtherViews.py?view=standard&confId=276820 for the conference that Matt mentioned. [10:42:45] Ewan Mac Mahon The key advantage of locking tapes in a safe for 30 years is probably that you'll be retired before anyone notices they don't work any more. [10:48:43] Jens Jensen Yes, the tape migration - we tend to do that every ~3 yrs anyway. [10:50:28] Jens Jensen Except when we need to buy new drives... [10:51:18] Steve Jones A very bright student could do some stats on this problems and write some guidelines. It's a good problem. [10:51:48] Steve Jones Many parameters and counter intuitive. [10:51:48] Matthew Viljoen Indeed - looks like a worth while part of the DPHEP initial investigation. [10:52:25] Ewan Mac Mahon I think the assumption should always be that any single thing will break, and the overall scheme should cope with that. [10:52:43] Ewan Mac Mahon I've said it about disks before - a disk is not a storage device, but a RAID array is. [10:52:55] Steve Jones Hm... [10:55:18] Ewan Mac Mahon *sends Jens a FOIA request for all the LHC data* [10:55:27] Steve Jones We never know what will be useful. One chap scrapped a laptop with bit coin worth £2m today! [10:55:31] Tom Whyntie You joke... [10:55:59] Jens Jensen https://www.gov.uk/government/publications?publication_filter_option=statistics [10:57:09] Steve Jones Sorry, £4m. [10:59:50] Ewan Mac Mahon On a personal note, my current desktop PC has, squirrelled away down a long directory path, some 512MB FAT filesystem images from an old PC I had that ran DOS as it's primary OS. [11:00:04] Ewan Mac Mahon They're now physically on a 4x1TB RAID array. [11:00:31] Steve Jones One could write book on that one topic easily. [11:00:40] Steve Jones Well, it wouldn't be easy to write....