Minutes of storage phone conference 03 May 2006

Attending:
	Edinburgh: Greig
	Lancaster: Matt, Brian
	RAL Tier 1: Derek
	RAL Storage: Jiri, Owen, Jens (chair+mins)

Apologies:
	Glasgow: Graeme


0. Review of actions (see below)

1. Deliverable status (again)

   Derek, Greig, and Owen are writing up the missing bits.
   Deliverable is a bit delayed, but Robin has been informed.

2. GridPP3 news or antinews

   Jens explained that he's writing up the case for support for
   storage in GridPP3.  The PMB are writing the GridPP3 proposal, and
   need input for storage - it is likely that funding will go down
   because storage is seen to become a "service".

3. DPM on NGS

   Jens reported that NGS sites are installing DPM to evaluate and
   report on it to the ETF - at the moment Belfast, Southampton, and
   IC (and Oxford, possibly).

   This is a Good Thing because GridPP to NGS convergence is seen as a
   Good Thing.  However, they are mostly failing; it seems they are
   trying to install 1.4.5 from LCG 2.7.0 on various non-SL systems,
   including SuSe and Solaris... (well, recompile on Solaris).

   One major issue appears to be that they all have 64 bit processors,
   and we have 0 experience with them apart from IC's dCache pool
   nodes [I checked with Alex from QMUL - they run their DPM on 32 bit
   machines]

   There is a specific mailing list for the SRM evaluation,
   ETF-SRM@jiscmail.  David McBride from IC is managing the list, and
   coordinating the evaluation effort.

   Jens also said he'd attend the ETF meeting this afternoon (Access
   Grid).

4. Storage for hepsysman?

   We had a lot of storage stuff at the last hepsysman in December,
   but then we had to meet SC4 goals and have SEs deployed at all
   sites.  Greig and Graeme will give talks, essentially
   updates/repeats of their HEPiX talks.  Jens will check with Pete
   whether there are any (other) storage issues that need covering
   (ACTION).

5. gLite 3.0 RC2-$ storage status
https://uimon.cern.ch/twiki/bin/view/LCG/Glite30IssueTracking

	DPM - believed to be 1.5.6.  Graeme sent a report before the
	meeting - _it_ doesn't work, but Graeme does.

	dCache - 1.6.6.5.  (Compare Tier 1 running 1.6.6.1)
	Greig says the client from 1.6.7 is out, with partial SRM2
	support (directories only AFAWK).  1.6.7 server hasn't been
	seen, but that one should have (most of) SRM2.1, and we're
	eagerly waiting to test it.  Need to follow up with DESY or
	FNAL again (ACTION).

6. Stress testing SEs?

   Work on concurrent reading and writing is ongoing (I saw a page to
   this effect in the wiki but now can't find it again).

   Jiri is working on parallel processes within S2 (the SRM2 testing
   program).

   At the moment, the best way to stress test an implementation is
   probably to plug it into the SC, but that won't necessarily work
   for all SEs.

7. dCache release procedure

   Since LCG is now not doing much - or any - testing, there is a need
   for us (GridPP, and DESY perhaps) to improve our release procedure,
   in particular the testing part.  ACTION Greig and Owen to suggest
   improvements.

8. Loose ends - except as covered above, there wasn't time.  We'll
   pick them up next time.

	Storage Items for Owen's dCache visit

	Filling in Greig's site template

	Status of testing other SRMs

	Status of testing CASTOR's SRM1.  Or SRM2.  Or, both.

	Poolgroups? VOMS support?

	FTS vs srmcp

	Reading to and writing from SRMs

	Review of issues for T2 workshop?

	Progress of storage security challenge

6. AOB

   How many files can an SE file.  If you ask for 500 files in one
   request, does it break?  Classic SEs did, but aren't SRMs.  Greig
   and Owen believe there is a setting in dCache to limit the number
   of files it will deal with.  Not known whether DPM has such a
   thing.  May need some testing.

   People should review the contents of the FTS vs srmcp page (ACTION).

   Brian asked whether dCache and DPM are using the same (in some
   sense) GridFTP.  They aren't really, dCache is using a Globus Java
   version, and DPM the Globus 2.4 C libraries.  They're doing work on
   transfer performance and are trying to "get rid of" TCP.

------------------------------------------------------------------------
ACTIONS

41	10/08/2005	Agree licence with DESY	Jens	Open		

No news.

53	12/10/2005	Find reasoanable % for SE uptime for SC4	Jeremy	Open		Reassigned. Follow up with GDB et al

No news.

54	02/11/2005	Report on performance/scalability with pools on WNs	Paul	Open		Progress. Will put in wiki. NB: dcache pools

No news, and nobody attended from Liverpool.

86	08/02/2006	Extend monitoring to do sites per VO and VOs per site	Greig	Open		

No progress since last meeting, but ongoing: Greig has already built
some visualisation for this information.  Jens suggested a
spreadsheet-readable format would be another useful way to publish
this information, and spreadsheet aficionados (like all of the PMB)
could then make their own visualisations too - Steve Lloyd had shown
such information at previous GridPP meetings.  It was agreed that this
was a good idea.

103	19/04/2006	Send dCache shopping list items to Owen/Jens, then prioritise	ALL	Open		

Ongoing, but now less urgent because Owen's trip has been postponed to
the end of May.

------------------------------------------------------------------------

NEW ACTIONS:

104	03/05/2006	Check storage requirements for May hepsysman	Jens	Open
105	03/05/2006	Re-poke DESY or FNAL about SRM 2.1 for dCache	Jens?	Open
106	03/05/2006	Suggest improvements to dCache testing/release procedure	Owen/Greig	Open
107	03/05/2006	Review FTS vs srmcp page	ALL	Open