Minutes of the storage EVO meeting, 06 Octal 2010

Present:
	Glasgow: David, Sam
	Edinburgh: Wahid
	Liverpool: Stephen, John
	Lancaster: Matt
	QMUL: Chris
	Sheffield: Elena
	Bristol: Winnie
	RAL: Brian, James, Jens (chair+mins)


There doesn't seem to be a NGS surgery today.


0.0 Storage at GridPP26 ?? (It's in April 2011, after the end of
   GridPP3.)

   Thanks to Wahid and Sam talking to Suzanne, who has now started
   planning for GridPP26, we can plan for a day or a day and a half of
   extra storage stuff.

0.4 Miscellaneous end-of-quartery items, like metrics and stuff.

    We got the blog numbers, but can we please remind people to blog?
    We know that it is being read :-)

0.7 Quick review of upgrades?  Anyone stuck?

    Discussion regarding whether StoRM 1.5 fixes the publishing
    problem?  It is believed to fix some of the problems, but not
    all.  Although the baseline release are 1.5.3 for StoRM, people
    should not upgrade till the problem is sorted.  Chris has 1.5 on
    SL4, but will not upgrade on SL5.

    1.5 has problems with publishing space (on Lustre), Chris will
    upgrade if experiments ask, but will not guarantee anything -
    we agreed that upgrading at this time is probably not a good idea.

1.0 The GDB on 13.10.2010 - overview of stuff we plan to do, if
   anything. (follow up on demos)

   http://indico.cern.ch/conferenceDisplay.py?confId=72063

   What is MSG?  It is a message queue which should ensure
   synchronisation between catalogues and SEs.

   Expecting demos of "solutions" from Jamboree this summer.

   Since there is lot of good stuff, is it worth listening in (but
   splitting it up so it becomes manageable) and reporting back to the
   group?  Our very own Graeme will of course be there and can also be
   persuaded to report.
   - Brian volunteered to listen (from somewhere) up to NFS4
   - Sam volunteered to go from ARC + 1 hour or so
   - Jens volunteered to report the remaining 1 hour, provided it
     doesn't clash with other stuff

   We don't necessarily need full coverage because slides will
   presumably be available but it may be useful to have someone listen
   in.

2.0 Overview of evaluations and their status and writeups, and next steps,
   if any.

   Next week will be devoted to the CHEP presentations.

   Sam's work with SSD is being written up simultaneously in the wiki,
   so progress can be tracked like that.

   Brian will write up the dCache install report, documenting the
   experiences so far.

3.0 Pre-review of events (as in: are there any obvious things we'll be
   interested in):

   CHEP  http://indico2.twgrid.org/conferenceTimeTable.py?confId=3

   OGF http://www.ogf.org/gf/event_schedule/index.php?event_id=17

   SC: http://sc10.supercomputing.org/schedule/schedule_all.php

3.7 EMI plans for data management

    Overview of items for (short-term) plan:
    * NFS 4.1
    * Commitment to standards
    * GLUE 2.0
    * "Harmonization" between the distributions (Unicore, gLite, ARC)
      - whatever that means.
    * Cloud interfaces via Unicore, cloud interface to dCache via CDMI
    * Catalogue SE integration
    * Further work on SRM, within the scope of OGF

4.9 AOB

    Long discussion followed on RFIO buffer sizes, and how to do the
    testing (eg with hammercloud), and impact on networks, and many
    other good things - see Wahid's presentation next week.

    Why did we even recommend large buffer sizes?  There was a time
    when the canonical file fortuitously fit into a large buffer; also
    more sensibly it may aid streaming of data - otoh Wahid is not
    seeing much impact on transfers.

[10:55:17] Queen Mary, U London London, U.K. joined
[10:55:17] Sam Skipsey joined
[10:55:17] Elena Korolkova joined
[10:55:17] Stephen Jones joined
[10:55:17] David Crooks joined
[10:55:17] Wahid Bhimji joined
[10:55:17] Matthew Doidge joined
[10:10:18] Jens Jensen http://indico.cern.ch/conferenceDisplay.py?confId=72063
[10:22:34] Wahid Bhimji http://hammercloud.cern.ch/atlas/10001349/test/
[10:22:44] Matthew Doidge 1 MB at Lancaster, but we haven't reviewed it in ages
[10:23:08] John Bland 0.5MB at Liverpool IIRC
10:56:49] Stephen Jones left
[11:01:42] Elena Korolkova left
[11:02:17] Jens Jensen The meeting has auto-extedned - you can keep going  
[11:03:33] Wahid Bhimji http://hammercloud.cern.ch/atlas/10001354/test/