Upload
nicholas-davidson
View
214
Download
0
Embed Size (px)
Citation preview
Overview of grid activities in France in relation to FKPPL
FKPPL Workshop
Thursday February 26th, 2009
Dominique Boutigny
FKPPL VO
Thursday February 26th, 2009Dominique Boutigny 2
Just 1 slide see Soonwook's talk…
The setting up of FKPPL VO has been the first practical action done within the framework of the VO
Idea: Setup a grid environment to allow the students attending the Seoul e-science and Grid school to practice on a real, full scale system Great
success !The VO has been setup very fast with good coordination between KISTI and CC-IN2P3Decision: end of July, first job in October
And the VO is actually used : 5000 jobs in 5 months and > 9000 kSI2k.hours
LCG Resources at CC-IN2P3
Thursday February 26th, 2009Dominique Boutigny 3
2006 2007 2008 2009 2010 2011 20120
5,000
10,000
15,000
20,000
25,000
0
10,000
20,000
30,000
40,000
50,000
60,000
Resource Deployment(Tier-1 + Analysis Facility)
Disk [TB] MSS [TB] CPU [k SI2000]
k S
I20
00
TB
2007 2008 2009 2010 2011 2012x 0.0
x 0.5
x 1.0
x 1.5
x 2.0
x 2.5
x 3.0
x 3.5
x 4.0
Planned annual increase rate of the installed capacity
(Tier-1 + Analysis Facility)
CPU Disk MSS
Roughly equivalent to 305 Servers (with 1TB
disks) or 34 racks
Disk storage deployment
Thursday February 26th, 2009Dominique Boutigny 4
X 6.7
Availability
Thursday February 26th, 2009Dominique Boutigny 5
Average: 92%
Reliability
Thursday February 26th, 2009Dominique Boutigny 6
01/01/200806/02/200813/03/200818/04/200824/05/200829/06/200804/08/200809/09/200815/10/200820/11/200826/12/200831/01/20090%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Average: 95%
Strengthening the system
Thursday February 26th, 2009Dominique Boutigny 7
Considerable efforts have been invested in 2008 / 2009 in order to strengthen the LCG computing infrastructure
Every piece is important – A single flaky component will ruin all the system We start to collect the fruits of this
ATLAS3000 jobs running in //
The weak point is clearly related to the SRM / dcache system which handles the data storage
Thursday February 26th, 2009Dominique Boutigny 8
CC-IN2P3 is committed to collaborate with KISTI in order to develop the ALICE Tier-2
LHC CPU consumption in 2008: > 17 M kSI2k.hours
Dominique Boutigny 9
clu
ste
r
Work on interoperability - JSAGA
Motivations for using several grid infrastructures:• increasing the number of computing resources available to user• need for resources with specific constraints
• super-computer• confidentiality• small overhead• interactivity
• availability, on a given grid, of:• the data• the software
Thursday February 26th, 2009
From Sylvain Reynaud
JSAGA
Thursday February 26th, 2009Dominique Boutigny 10
WMS
inputdata
SRM
GridFTP
WS-GRAMLCG-CELCG-CE WS-GRAMfir
ewal
l
jobdesc.
gLiteplug-ins
Globusplug-ins
JSAGA
job
staginggraph
dele
gate
sel
ectio
n
& fil
es s
tagi
ng
job
OPlastEGEE
hide
infr
astr
uctu
res
hete
roge
neity
(e.g
. EG
EE
, OS
G, D
EIS
A)
hide
mid
dlew
are
hete
roge
neity
(e.g
. gLi
te, G
lobu
s, U
nico
re)
JDL RSL
From Sylvain Reynaud
Dominique Boutigny 11
Ready-to-use software, adapted to targeted scientific field
Hide heterogeneity between grid infrastructures
Hide heterogeneity between middlewares
As many interfaces as ways to implement each functionality
As many interfaces as used technologies
Applications
SAGA
end userapplication developer
plug-ins developer
core engine+ plug-insJSAGA
jobscollectionJSAGA
Thursday February 26th, 2009
From Sylvain Reynaud
A JSAGA application
Thursday February 26th, 2009Dominique Boutigny 12
JUX is a file explorer designed to be independent of– Operating System
• tested on Windows, Scientific Linux, Ubuntu, Mac
– Data management protocol• tested with gsiftp, srb, irods, http, https, sftp, zip,
(srm)– Security mechanism
• tested with GSI, VOMS, Login/Password, X509, SSH
full javacode
JSAGA
Pascal Calvat + Sylvain Reynaud
For instance, it is possible to interactively handle files stored in SRM / dcache from my own laptop and to move them to another data storage system managed by another Grid middleware
Accessing parallel computer through the EGEE middleware
Thursday February 26th, 2009Dominique Boutigny 13
Some applications need to run on parallel computer:• Molecular Dynamics within WISDOM• Lattice QCD• Some astroparticles applications• …
EGEE middleware has been mainly designed to address jobs to serial computer farmsUsing parallel computers would require to be able to characterize the parallel nodes within the Information System
Very relevant in the framework of FKPPL
Parallel computers at CC-IN2P3
Thursday February 26th, 2009Dominique Boutigny 14
CC-IN2P3 operates a small parallel farm:232 CPU-cores connected in Gigabit Ethernet
This farm will be upgraded this year to~1000 CPU coresLow latency network (probably Infiniband)
Due to modern CPU design constraints (many cores per chip), using parallelism even in HEP applications will become unavoidable
I consider that gaining expertise in this area is crucial for CC-IN2P3
KISTI has this expertise !
Analysis interactive platform
Thursday February 26th, 2009Dominique Boutigny 15
This year CC-IN2P3 will build a powerful analysis interactive platform
Fast event filtering - Typically read and process AOD at 1 kHz for 50 users in parallel
Root analysis
The architecture will be based on PROOF + probably xrootd, but other storage system will be considered
A prototype will be setup in the coming weeks with existing hardware in order to validate the architectureThen we will build a full scale system for LHC startup
Thursday February 26th, 2009Dominique Boutigny 16
Analysis interactive platform
ALICE is very enthusiastic to get such a system at CC-IN2P3 which will complement the CERN Analysis Facility
We will easily get user applications to test the system
Also in contact with René Brun and PROOF development team in order to setup something really powerfulBalance between RAM – SSD and HDD
This is something that I propose to consider within the framework of FKPPL