Minutes of NAF User Committee meeting from 13.05.2009 
-----------------------------------------------------

Present/Phone: Steve Aplin (ILC), Wolfgang Ehrenfeld (ATLAS), Andreas
Gellrich (NAF), Andreas Haupt (IT), Yves Kemp (IT), Kai Leffhalm
(NAF), Birgit Lewendel (IT), Niels Meyer (ILC), Hartmut Stadie (CMS),
Jan Erik Sundermann (ATLAS), Alexey Zhelezov (LHCb)


1. News from the chair

Nothing to report.


2. Action items:

Only items with updates, see action item list for complete list of
items.

 - Test NAF SL5 setup for software:

ATLAS: athena analysis job is running on SL5, compiling on SL5 using
SL4 libraries needs to be tested

CMS: compiling and running is fine

LHCb: should work, will adapt in time of switch over

ILC: compiling on SL5 is fine, running not yet tested

NAF: will prepare a schedule for switch over

Item will stay open. More feedback needed/expected by the experiments.


 - Port issue from CVS/kerberos with D4:

DESY D4 adjusted the firewall settings using a list of hosts hosting
CVS server at CERN. Works now as at CERN. Closed.


 - Feedback to automatic proxy prolongation:

CMS did some tests and announced it to their users. No feedback yet.
Action item open til next meeting. 

 - Transfer guidelines/matrix:

See operations report, page 4, for some discussion.

 - Working example for gsiscp between two VOs and working example for
   rsync with gsissh

This works as expected but is depreciated for medium and large data
volumes as gsi based transfers have to go to the VOs login
hosts. Bandwidth is limited there. For more details and some example
see operations report, page 4. Closed.


3. Status report:

Status report was given by Kai. See the agenda for the report.
Below a few highlights from the discussion are listed:

The information flow from the NAF admins to the Users when the cooling
problem occurred was criticised. More information is needed. It was
suggested that after some time (e. g. one hour) the NAF admins will
give an estimated of the expected downtime length. This helps the
users to better judge if they should wait or do something else.

The experiments should advertise the German VO role better,
e.g. /atlas/de. /cms/dcms, /ilc/de and /calice/de. This is the only
way to do accounting on the NAF Grid resources. This is particularly
important for the NAF funding.

It was noticed by the NAF admins that most of the batch jobs run less
than one hour. This is different from the user behaviour a year
ago. Slots in the batch system below one hour are mainly for PROOF
sessions and are implemented using overloading of the batch
nodes. This is okay from time to time but not permanently. 

After some discussion, the idea is to remove the overloading for jobs
below one hour and switch emphasis from 12 hour analysis jobs to one
hour analysis jobs. The NAF admins will discuss this internally and
report back to the NUC.

CMS requested a more detailed presentation of the job waiting
time. This includes per VO and per queue time (<15min, 15min-1h,
1-12h, >12h).


4. LHCb NAF usage: MC production

LHCb tested the NAF infrastructure producing some MC samples for a
student as a test case. It worked quite well, more suitable than the
Grid. For more details see Alexeys report.

Better availability of Grid resources would be possible through the
/lhcb/de VOMS group, which is not supported by LHCb. LHCb should get
in contact with the IT people from Zeuthen to check what can be done
to improve this.


5. AOB

Nothing.