Transcript Palermo

INAF experience in Grid
projects
C. Vuerli, G. Taffoni, V. Manna,
A. Barisani, F. Pasian
INAF – Trieste
Overview
• The GRID.IT Project
• DRACO and the VO INAF
• Planck Simulations in EGEE
• New powerful Grid infrastructures
– TriGrid Virtual Laboratory
• Training and Dissemination
Wed 10 May 2006
2
GRID.IT Project
• The GRID.IT Project
– Application 1
• Accessing Databases through the GRID
• Processing Data extracted from Databases
• Ingesting results of Data Processing in Databases
– Application 2
• Porting in GRID the VST Pipelines
– Application 3
• Monitoring of Astronomical Instrumentation through the
GRID
Wed 10 May 2006
3
GRID.IT / Application 1
• Problem
– The middleware based on Globus 2.4 didn’t
offer any mechanism to access databases
• Solutions
– Evaluation of a number of possible solutions
• Spitfire
• OGSA-DAI
• G-DSE
Wed 10 May 2006
4
GRID.IT / Application 1
• Spitfire
– Not suitable for us (poor functionalities)
• OGSA-DAI
– WS-based: we don’t want to merely access data. We
want databases integrated in Grid and look at DBMS
machines as computational machines
• G-DSE
– Databases as embedded resources in GRID
Wed 10 May 2006
5
GRID.IT / Application 3
• Past experience in developing Telescope
and instrumentation control systems and
participation to EU pilot projects to
remotely monitor and control them
– We want to go ahead with this experience
and use GRID for this purpose
– First simple monitoring prototype: tests
carried out at Trieste and Naples
Wed 10 May 2006
6
GRID.IT / Future evolution
• G-DSE + QE
– G-DSE prototype engineering
– Design, implementation and test of the QE
– G-DSE + QE in production
• G-ICS + IE
– Design, implementation and test of the G-ICS
and of the IE
– G-ICS + IE in production
Wed 10 May 2006
7
In production…but where ?
• G-DSE + QE
– All INAF Databases in GRID with G-DSE
– Other scientific communities are testing it : INFN,
Bioinformatics, Archaeologists, …
– Supporting Planck simulations in future releases
– Projects in preparation to port it on EGEE and DEISA
• G-ICS + IE
– Close contacts with GridCC
– Strong interest for instruments and sensors in Grid
• ICTP and INAF Naples : monitoring and controlling networks
of sensors via Grid
Wed 10 May 2006
8
DRACO
• A Grid portal thought and designed for astrophysical
•
applications
Porting astrophysical applications in GRID
– Definition of astronomical data analysis workflow on a serviceoriented Grid architecture using BPEL (Manna et al.)
– Data Mining and Massive computing for GRID at Naples-Salerno
Unit
– Astronomical Data Reduction in Data Grid
– Various kinds of evolutionary models
– ACME for GRID
• The INAF VO is now growing
Wed 10 May 2006
9
Planck simulations in EGEE
• Planck simulations is an officially
supported pilot EGEE application since
November 2004
• Planck application also in EGEE-2
• A Planck VO, led by INAF Trieste, is in
place. It groups together Grid sites located
in six European countries
Wed 10 May 2006
10
Planck simulations in EGEE
• First experiments to port the Planck simulations
in Grid at INAF-Trieste
– Partial Planck simulation pipelines in June 2004 using
the GRID.IT Grid infrastructure
– Simulation of the whole Planck mission (two complete
sky surveys) in Grid and comparison of results with
those got on a single dual-CPU workstation
– New bricks of the Planck pipeline in Grid in the next
months
• WG8 meeting at MPA (Germany) in June 2006
Wed 10 May 2006
11
Planck simulations in EGEE
• Planck simulations didn’t make use of databases
until now, but…
– They could be profitably integrated and used in the
near future
– Ground based test and calibrations of the spacecraft
and of the instruments are now in progress
– Data collected in this phase must be saved
somewhere and used during the operative phase of
the mission
• …therefore this is a new test-bed for G-DSE
Wed 10 May 2006
12
TriGrid Virtual Laboratory
Wed 10 May 2006
13
TriGrid Virtual Laboratory
• Catania Pole:
– ~100 machines dual CPU 32/64 bit. Aggregated power of 300
kSpecInt2000.
– 30 TB of storage EIDE-to-SCSI or FC-to-SATA in RAID 0,1,5
– 20 working stations for development/porting of applications
• Messina Pole:
– ~80 machines dual CPU 32/64 bit. Aggregated power of ~240
kSpecInt2000.
– 20 TB of storage EIDE-to-SCSI or FC-to-SATA in RAID 0,1,5
• Palermo Pole:
– ~30 machines dual CPU 32/64 bit. Aggregated power of ~100
kSpecInt2000.
– 10 TB of storage EIDE-to-SCSI or FC-to-SATA in RAID 0,1,5
Wed 10 May 2006
14
Training and dissemination
• The knowledge of GRID must grow both
at national and European level
– It is necessary to train Grid managers able to
install, configure and maintain Grid sites
– It is necessary to train astronomers to
illustrate them this new technology and make
them aware of what they can profitably do
with the Grid for their everyday work
Wed 10 May 2006
15
Training and dissemination
• How to train ?
– Meetings and workshops both at national and
European level
• Theoretical lessons
• Practical training
– Astronomers bring applications of interest and try to gridify
them with our help
– On site dedicated workshops
– Grid schools
– Coordination with Working Groups having the same
objectives (the Planck Working Group 8)
Wed 10 May 2006
16
The End
Thank you for your attention
Wed 10 May 2006
17