Notes of the CCRC'08 planning conference call of 15.10.07
Present:
WLCG Service Coordination: J.Shiers (chair - JDS) H.Renshall (notes - HR)
Tier0: M.Coelho (MC) WLCG/LHC-Ops link: absent MB: A.Aimar (AA)
ATLAS: J.Shank (JS) CMS: M.Kasemann (MK) ALICE: L.Betev (LB)
LHCb: N.Brook (NB) ASGC: J.Shih (JS) BNL: M.Ernst (ME)
CNAF: L.dell'Agnello (LA) FNAL: absent FZK: A.Heiss (AH)
IN2P3: F.Hernandez (FH) NDGF: M.Wadenstein (MW) NL-T1: M.vdSanden (MS)
PIC: G.Merino (GM) RAL: A.Sansum (AS) TRIUMF: R.Tafirout (RT)
Resulting actions for next meeting (22 October):
1) Experiments to quantify their targets for the CCRC'08 runs as shown in
the agenda next presentation slide - CMS Basic Scaling Items to check
in CSA08 - appropriately modified to their computing model.
2) A.Aimar to present (very) preliminary CCRC'08 milestone plans.
3) WLCG Service Coordination to gather the planned release dates of the
required new middleware.
Notes:
JDS went over his presentation 'Tasks for first phone-call' (attached to
the agenda) with various discussions as he went along.
As regards the CMS Proposed Schedule (slide 2) NB said that
LHCb have no strong plans for this period (Feb-May 2008) and
will simply use the CCRC'08 challenges as part of their FDR (Full-scale
Dress Rehearsal) activities which would be at their full p-p rates.
The CMS timeline (slide 4) provoked two points. For ALICE LB said that the
proposed schedule looked ok but asked if their DAQ was expected to reach
their p-p goals in February since they may not have all detector components
contributing. JDS replied they should use whatever they have at the time.
There was discussion of coordinating these activities with experiment
software releases and it was agreed it would be useful for experiments
to prepare a similar timeline to that of CMS as shown on slide 4.
JDS went over the Explicit Requirements (slide 6) in reverse order.
On the issue of demonstrating the scalability of the conditions DB NB
said LHCb had tested this at some sites and he will check for the others.
JDS pointed out that ATLAS will have a session on this in their next
computing workshop in the week of 22 October.
Under the requirement of CMS to have commissioned links it was pointed
out that those that drop back below threshold immediately trigger a
recommissioning program.
For the gLite 3.1 VO Box requirement of ALICE LB said this needs more
rigour in certification and JDS promised to raise this with the GD group
EMT (Engineering Management Taskforce).
The issue of the availibility of the gLite 3.1 WMS under slc4 (currently
in integration) also affects ALICE. The slc3 version is in production as
a backup and LB would like the sites to give their views on this. LC asked
how to cope with slc3 services in general and JDS remarked that server
side machines at CERN will stay on slc3 until ready to migrate.
On pilot jobs the WLCG management will propose to the collaboration board
that sites accept pilot jobs using one of the GLexec mechanisms.
For Read-only LFC replicas only packaging and release remain to be
completed.
JDS proposed the ALICE issue of xrootd interfaces at all ALICE sites be
referred back to the MB. LB though this was premature and preferred to
hear first from the CERN CASTOR integration team (dcache is ready).
JDS reminded that roll-out of SRM 2.2, the last issue, remained the top
priority. MK thought that upgrades of local MSS systems (e.g. to dcache
1.8) should also appear as an explicit requirement.
There was then some discussion of the Implicit Requirements (slide 11).
JDS would like to know what is the effect on an experiment of degraded
services, e.g. if a conditions D/B went down during a weekend, so we
would know where to put effort to make imrovements. MC thought we should
also define thresholds within services in order to trigger remedial actions.
Discussion then moved to the Issues on the last slide.
There was a concensus that the quarterly per site resource requirements
tables were sufficiently detailed and HR said he was busy preparing them
for the first half of 2008 but needed plans from the Tier 1 sites, in
particular to know what resources would be available in February as
this would limit what the experiments could do.
JDS speculated that not all experiments would follow the staged buildup
of the February challenge as proposed by CMS on slide 2.
JDS thought achieving the explicit requirements was feasible but we will
need to know the sites upgrade plans to reach these requirements. LB
reminded that ALICE can work with hybrid setups - slc3 and slc4 services
and gLite 3.0 and 3.1 WMS.
The last question was if Gant charts might be useful as a planning tool.
AA thought not as it was hard to add useful text comments and suggested
instead a format like the MB milestones. He would make a first attempt.
The closing discussion was on the date and preparations for the next
meetings. It was agreed for the moment to continue with weekly meetings
so the next will be on 22 October at 17.00 CET (the agenda is in place).
We would later use the CCRC'08 milestones plan to trigger any meetings in
addition to those planned during the pre-gdb days. For the 6 November
pre-gdb it was agreed to meet in the afternoon to help USA members.
The main item for 22 October would be to review experiment targets to be
achieved in the February and May challenges - similar to that of CMS
(slide 16) with quantified numbers such as the Tier-0 reconstruction rate
and the number of daily job submissions to Tier-1 and Tier-2 sites so
experiments are asked to prepare such input. In addition WLCG Service
Coordination will gather the planned release dates of the required new
middleware.
It was decided to look at the resource requirements to meet the experiment
targets in the following weeks meeting. The week after that would be a
pre-gdb.
Finally MK asked if the mailing list for this meeting was closed or open
as they wished to add their Tier-2 representatives. JDS said that he and
HR had to approve joining requests and that this would be done.