May 2004Sverre Jarp1 Preparing the computing solutions for the Large Hadron Collider (LHC) at CERN...
-
Upload
clarence-hodges -
Category
Documents
-
view
217 -
download
1
Transcript of May 2004Sverre Jarp1 Preparing the computing solutions for the Large Hadron Collider (LHC) at CERN...
May 2004 Sverre Jarp 1
Preparing the computing solutions
for the Large Hadron Collider (LHC) at CERN
Sverre Jarp, openlab CTOIT Department, CERN
May 2004 Sverre Jarp 2
Short overview of CERN
May 2004 Sverre Jarp 3
150m depth150m depth150m depth150m depth
Accelerators and detectors in underground tunnels and caverns
May 2004 Sverre Jarp 4
CERN in numbers Financed by 20 European countries
Special contributions also from other countries:
USA, Canada, China, Japan, Russia, etc.
1000 CHF (650 M€) budget to coveroperation + new accelerators
2,200 staff (and diminishing)
6,000 users (researchers) from all over the world
broad visitor and fellowship program
May 2004 Sverre Jarp 5
CERN User Community
Europe: 267 institutes, 4603 usersElsewhere: 208 institutes, 1632 users
May 2004 Sverre Jarp 6
Computing at CERN
May 2004 Sverre Jarp
reconstruction
simulation
analysis
interactivephysicsanalysis
batchphysicsanalysis
batchphysicsanalysis
detector
event summary data
rawdata
eventreprocessing
eventreprocessing
eventsimulation
eventsimulation
analysis objects(extracted by physics topic)
event filter(selection &
reconstruction)
event filter(selection &
reconstruction)
processeddata
Data Management and Computing for Physics Analysis
May 2004 Sverre Jarp 8
High Energy Physics Computing Characteristics
Independent events (collisions of particles) trivial (read: pleasant) parallel processing
Bulk of the data is read-only versions rather than updates
Meta-data in databases linking to “flat” files Compute power measured in SPECint (not SPECfp)
But good floating-point is important Very large aggregate requirements:
computation, data, input/output Chaotic workload –
research environment - physics extracted by iterative analysis, collaborating groups of physicists
Unpredictable unlimited demand
May 2004 Sverre Jarp 9
SHIFT architecture (Scalable Heterogeneous Integrated Facility)
Tapeserver
Diskserver
Batchserver
Interactiveserver
BatchanddiskSMP
Network- Ethernet
AFS
In 2001 SHIFT won the 21st Century Achievement Award issued by Computerworld
DiskserverDiskserver
BatchserverBatchserver
TapeserverTapeserver
May 2004 Sverre Jarp 10
CERN’s Computing Environment (today)
High-throughput computing (based on reliable “commodity” technology)
More than 1500 (dual Xeon processor) PCs with Red Hat Linux
About 3 Petabytes of data (on disk and tapes)
May 2004 Sverre Jarp 11
IDE Disk servers Cost-effective disk storage:
< 10 CHF/GB (mirrored)
May 2004 Sverre Jarp 12
The LHC Challenge
May 2004 Sverre Jarp 13
Large Hadron Collider A completely new particle
accelerator The largest superconductor installation in the
world Same tunnel as before; 27 km of magnets Super-fluid Helium cooled to 1.9°K Two counter-circulating proton beams in a
field of 8.4 Tesla
Collision energy: 7 + 7 TeV
Simulation tool:Sixtrack
May 2004 Sverre Jarp 14
CMSATLAS
LHCb
Accumulating data at 10 Petabytes/year (plus replicated copies)
Requirements:
Storage – Raw recording rate: up to 1 GB/s per experiment
2 Petabytes of disk/year
Total HEP processing needs – 50,000 (100,000) of today’s fastest processors
The Large Hadron Collider (LHC) has 4 Detectors:
May 2004 Sverre Jarp 15
The Large Hadron Collider (LHC) goal:
All charged tracks with pt > 2 GeV
Reconstructed tracks with pt > 25 GeV
(+30 minimum bias events)
Find new
physics,
such as the
Higgs
particle, and
get the
Nobel price !
selectivity: 1 in 1013 - 1 person in a thousand world populations
May 2004 Sverre Jarp 16
LHC Computing Grid Project
Goal of the project:To prepare, deploy and operate the computing environment for the experiments to analyse the data from the LHC detectors
Phase 1 – 2002-05development of common applications, libraries, frameworks, prototyping of the environment, operation of a pilot computing service
Phase 2 – 2006-08acquire, build and operate the LHC computing service
The Grid is
just a
tool to
wards achieving this goal
May 2004 Sverre Jarp 17
RAL
IN2P3
BNL
FZK
CNAF
PIC ICEPP
FNALComputing Model (simplified!!)
Tier-0 – the accelerator centre
Filter raw data Reconstruction summary data (ESD) Record raw data and ESD Distribute raw and ESD to Tier-1
Tier-1: Permanent storage and management of
raw, ESD, calibration data, meta-data, analysis data and databases grid-enabled data service
Data-heavy analysis Re-processing raw ESD National, regional support “online” to the data acquisition
process high availability, long-term
commitment managed mass storage
Tier-2: Well-managed disk storage –
grid-enabled Simulation End-user analysis – batch and
interactive High performance parallel
analysis (PROOF)
USCNIKHEFKrakow
CIEMAT
Rome
Taipei
TRIUMF
CSCS
Legnaro
UB
IFCA
IC
MSU
Prague
Budapest
Cambridge
Tier-1small
centres
Tier-2
desktopsportables
May 2004 Sverre Jarp 18
RAL
IN2P3
BNL
FZK
CNAF
USC
PIC ICEPP
FNAL
NIKHEFKrakow
Taipei
CIEMAT
TRIUMF
Rome
CSCS
Legnaro
UB
IFCA
IC
MSU
Prague
Budapest
Cambridge
Processing M SI2000**
Disk PetaBytes
Mass Storage
PetaBytes
CERN 20 5 20
Major data handling centres
(Tier 1)45 20 18
Other large centres (Tier 2)
40 12 5
Totals 105 37 43
** Current fast processor ~1K SI2000
Current estimates of Computing Resources needed at Major LHC Centres
First full year of data - 2008Data distribution~70 Gbits/sec
May 2004 Sverre Jarp 19
LCG Basics Getting the data from the detector to the grid
requires sustained data collection and distribution keeping up with the accelerator
To achieve the required levels of performance, reliability, resilience
at minimal cost (people, equipment) we also have to work on scalability and performance
of some of the basic computing technologies: cluster management mass storage management high performance networking
Workshop 3
Workshop 1
Workshop 2
Workshop 5
May 2004 Sverre Jarp 20
SW Rep
Fabric Automation at CERN
Node
CfgCache
SWCache
SPMA
SWRepCDB
CDBOraMon
NCMMSA
SMS
HMS
LEMON
LEAF
ConfigurationInstallation
Fault & hardware
Management
Monitoring
Includes technology developed by DataGrid
May 2004 Sverre Jarp 21
WAN connectivity
Itanium-2single stream:
5.44 Gbps1.1 TB in 30 mins
We now have to get from an R&D project (DATATAG) to a sustained, reliable service – GEANT, ESNET, ..
Microsoft enters the stage:
Multiple streams: 6.25 Gbps
20 April 2004
May 2004 Sverre Jarp 22
Preparing for 2007
The LCG installation is on a tight schedule
Due to the need for deployment and development in parallel
2004
2005
2006
2007
first data
Initial service in operation
Decisions on final core middleware
Demonstrate core data handling and batch analysis
Installation and commissioning
May 2004 Sverre Jarp 23
CERN openlab
May 2004 Sverre Jarp 24
openlab: The technology focus of CERN/IT
Industrial Collaboration:
Enterasys, HP, IBM, and Intel and Oracle are our partners
Voltaire (with Infiniband switches) just joined
Technology aimed at the LHC era: Network switches at 10 Gigabits ~ 100 rack-mounted HP servers 64-bit computing: Itanium-2
processors StorageTank storage system
w/28 TB ~1 GB/s throughput
May 2004 Sverre Jarp 25
64-bit porting status
Ported: Castor (data management subsystem)
GPL. Certified by authors. ROOT (C++ data analysis framework)
Own license. Binaries both via gcc and ecc. Certified by authors.
CLHEP (class library for HEP) GPL. Certified by maintainers.
GEANT4 (C++ Detector simulation toolkit) Own license. Certified by authors.
CERNLIB (all of CERN’s FORTRAN software) GPL. In test.
Zebra memory banks are I*4
ALIROOT (entire ALICE software framework) LCG-2 software from VDT/EDG
GPL-like license.
Being ported: CMS ORCA (part of CMS framework)
May 2004 Sverre Jarp 26
CERN“Where the Web was born…”®
CERN is busily preparing for the first arrival of LHC data in 2007 New and exciting technologies are
needed to manage the data Seamlessly, around the globe Together with our partners (EU, industry, other
Physics Labs, other sciences) we expect to come up with interesting proofs-of-concept and technological spin-off !
High Throughput Computing is “on the move” !
People Motivation
Technology Science
Innovation