Fit and Healthy? Glenn Patrick Rutherford Appleton Laboratory GridPP26 29th March 2011 or just lethargic?

Download Report

Transcript Fit and Healthy? Glenn Patrick Rutherford Appleton Laboratory GridPP26 29th March 2011 or just lethargic?

Fit and Healthy?
Glenn Patrick
Rutherford Appleton Laboratory
GridPP26
29th March 2011
or just lethargic?
2
H1: UK Monte-Carlo
c/o Dave Sankey & Bogdan Lobodzinski
GGUS TICKETS
01.01.2010 – 01.04.2011
UKI-NORTHGRID-MAN-HEP
RAL-LCG2
UKI-LT2-BRUNEL
UKI-NORTHGRID-LANCS-HEP
UKI-SOUTHGRID-BHAM-HEP
UKI-LT2-QMUL
UKI-LT2-RHUL
UKI-NORTHGRID-LIV-HEP
UKI-LT2-UCL-CENTRAL
UK TOTAL (40%)
2
6
7
3
2
6
4
2
1
33
Usually aborted jobs or unavailable SE.
Most tickets solved by action in authentication/authorisation area.
3
http://www-h1.desy.de/h1/www/h1mc/mts.html
H1: UK Monte-Carlo
25% of all H1 jobs from 12 UK sites.
UK-DESY transfer rate only ~500kB/sec (averaged over 2 months).
4
Linear Collider
Silicon Detector (SiD) Design Study –
tracking and calorimetry.
ILC ~500 GeV.
LOI published 31 March 2009
c/o Jan Strube
Multi TeV (~3 TeV) e+e- collider
Two beam accelerator
5
Linear Collider: Status
Obviously using DIRAC
as well as Ganga
CLIC is in the middle of a
conceptual design report (CDR)
5 different benchmarking
channels at 3 TeV + 1 at 500 GeV
Software needed a lot of work to
move from 500 GeV to 3 TeV
--> delays in the start of
production
312 background events / signal
events !!!
6
Linear Collider: Next Steps
Current Storage used:
• RAL-SRM : 1,513,322,629,922
• KEK-SRM : 27,825,900,796
• CERN-SRM : 15,468,811,712,135
• IN2P3-SRM : 1,362,953,567,143
Total : 18,372,913,809,996
Background merging step planned to be done at RAL.
-> Significant increase expected, but no need to increase T1
allocation foreseen
7
Linear Collider: Long Term
CLIC CDR due in Fall 2011
• Production just about to hit the hot phase.
ILC detectors to write Detector Baseline Document:
• Due to end 2012.
• Last time: Production mostly at SLAC Batch farm.
• This time: Expected to exclusively use Grid
(expected merger of DESY ILC VO w/ Fermigrid ILC VO would
provide relief for Tier1).
• Total scale ~ 1/4 of CLiC total.
• Tier1 allocation at the current level should be sufficient.
 But limited UK manpower in the future
(both Jan and Marcel depart).
8
SuperB: Newly born?
Approved by Italian Government
14/15 December 2010.
Three sites under consideration:
two near Rome(LNF and Tor
Vergata) or green field site.
• Re-use components from PEP-II and BaBar.
• UK interested in building the Silicon Tracker.
• MAPS pixel sensors and modules (~1m2 of silicon).
• Opportunities also on accelerator side.
• UK Grid resources for computing model & leadership.9
SuperB: Computing
Computing guesses 2020
(BaBar extrapolation)
Nominal luminosity (15 ab-1/year)
TAPE 100 PB
DISK 50 PB
CPU 1700 kHEPSPEC06
• Now - Fast Grid simulation for physics and detector studies.
• Run at QMUL, RAL and Oxford.
• Produced ~20% of Monte-Carlo.
• Babar code converted to run on Grid, but needs recoding.
• In next 2 years ramp up detector simulation.
• Want to get accelerator studies into computing model.
“Our efficiency is infinite since we are producing events
with no resources”.
c/o Fergus Wilson
10
SNO+ Neutrino Detector
• SNO+ is a multi-purpose liquid
scintillator neutrino detector.
• Like SNO, but with heavy water
returned and replaced.
• H2O in 2012 and liquid
scintillator in 2013.
• More interaction light means
more data. Need to store and
process many TB of data.
• Monte-Carlo is CPU intensive
due to many photons that need
to be tracked.
• Looking to use Grid in Europe:
Oxford, Sussex, QMUL,
Liverpool and Lisbon.
• Early days – VO just set up.
• Probably using T2 sites for MC
production and storage.
c/o Jeanne Wilson
11
T2K
Super-Kamiokande
Kamioka mine (1000m deep)
50 kton water Cherenkov
Long-baseline neutrino
oscillation experiment
(295km)
Measure oscillation
of νµ to νe
Neutrino beam generated using 50 GeV proton
synchrotron at J-PARC facility in Tokai.
12
T2K: A few other problems...
via Geoff Pearce
13
T2K: Data Model
100% Raw Data at RAL, TRIUMF (IN2P3) sites.
Weighted percentage at T2 sites dependent on resources.
FTS server set up at RAL and UK channels working.
c/o Ben Still
14
T2K: Grid Progress
• Now starting to heavily use the Grid for processing as well as
distribution and storage of data.
• System in place to distribute new data from RAL and TRIUMF
Tier 1 centres to Tier 2 sites in UK, Spain (Barcelona and
Valencia) and France (IN2P3). Done via FTS with channels
hosted at RAL.
• Issues with just using lcg-utils to copy data from Japan.
• Storage snapshot. RAL Tier 1 ~109TB. UK Tier 2 sites ~1- 40TB.
• Soon entering largest data processing phase – Grid will play
major role.
• Team of 3 post-docs working on Grid related issues: Ben Still
(QMUL), Gustav Wikstrom (Geneva) and Jon Perkin (Sheffield).
c/o Ben Still
15
NA62
K   


NA62-II (2013-2014): UK – Birmingham, Bristol, Glasgow, Liverpool.
• Short test end of 2012.
• Large MC run in Autumn 2011.
• VO created in UK, enabled at Glasgow.
c/o Dave B.16
• Grid interface (Janusz).
MICE
MICE magnet delays, etc. Rescheduling.
Some tests/runs in 2011.
Next “Step” in 2012.
Custodial copy of raw data (2 copies) at UK T1.
Important that this is efficient!
c/o Dave Colling
17
PhenoGrid: Health Warning!
1. Unlike large VOs, there is no dedicated support staff. Chasing problems
takes academic time away from research.
2. Don’t have the ability to submit team tickets. Appears tickets get low
priority because they are considered individual user (not VO) problems.
Let to a number of issues in last 6 months taking a long time to resolve
and Grid being unusable for much of that period. Hoping this will improve
after interaction with Jeremy.
3. Stability of Grid service is still dismal. Too many potential sources of point
failure leading to a total up time of the Grid being well below 50% in our
opinion – rather than anticipated high 90%+.
4. Grid middleware considered to be a failure on a scale that would
embarrass a government IT project. Still inconsistent, unstable and
unreliable after ~10years of development. Fails the “Daily Mail” test.
c/o Peter Richardson
18
Little Sign of Life?
“Maybe he’s dead, Jim”
19
Not to Forget...
MINOS: Nick West and Phillip
Rodrigues have left.
Talked to Alfons Weber (UK
Spokesman) – now little UK effort
and expect limited UK Grid use.
Some legacy issues over storage
(NFS, etc). Fire in Soudan mine!
Future neutrino experiment (NOvA)?
SuperNemo: Gianfranco Sciacca has
left. Not sure of status – no update on
“health”.
CDF and DZERO: Little UK Grid use
now (none at T1). I did not consult.
20
21