NATIONAL INSTITUTE FOR LASER, PLASMA ANDRADIATION PHYSICS

Institute of Space Sciences

P.O. Box: MG-23, RO 077125 Bucharest ROMANIATel./Fax (4021) 457.44.71

http://www.spacescience.ro

ISS-AliEn and ISS-gLite

Adrian Sevcenco

RO-LCG 2011 WORKSHOP

Applications of Grid Technology and High Performance Computing in Advanced Research

Main directions for R & D activity

• Space physics:– cosmic rays, – High Energy Physics– nuclear astrophysics and particles – space plasma and magnetometry

• Cosmology • General theoretical and mathematical physics• Gravitation and Microgravitation • Space technology:

– engineering for space research and applications

ISS Computing Facilities

AliEn cluster – ALICE dedicated cluster

gLite cluster – gLite middleware cluster; VO ALICE so far

RoSpaceGrid cluster – Space Plasma and Magnetometry Group (GPSM) - ISS

PlanckGrid cluster -

392 CPU740 GB memory

140 TB storage space 30% of total Romanian Tier 2 Federation

2% of ALICE computing (last 1 year period)

ISS COMPUTING FACILITIES

ISS-AliEn

1 year cluster activity

ISS-AliEn

ISS-gLite

Storage status

RoSpaceGRID - PECS-ESA

Romanian GRID middelware repository for Space Science Applications

The RoSpaceGRID framework is an implementation of GRID which consists of clusters, labs with distributed computing resources all around the world, where computer activities can be submitted for scheduled execution. The RoSpaceGRID is deployed in Institute of Space Science (ISS) for PLANCK, CLUSTER II and VENUS EXPRESS space missions.

This project will be established for the benefit and in cooperation with the scientific community of the three missions mentioned above.

R&D GRID

Future developments

• Main data traffic : WNs <----> local storage

• Change local network infrastructure to 10 GigE (dec 2011 – jan 2012)

CUDA parallel computing

Compute Unified Device Architecture   NVIDIA General purpose parallel computing for GPU Requirements:

ISA(Instruction Set Architecture) NV driver CUDA Toolkit (CUDA SDK)

CUDA parallel computing

Execution Model

CUDA parallel computing

Performance optimizations

Expose as much parallelism as possible

Optimize memory usage to achieve maximum memory throughput

Optimize instruction usage to achieve maximum 

instruction throughput

Maximize occupancy to hide latency

Promising Future New building:

• National Center for Space Science and Technology

•Description:•Basement - Data Center, CleanRoom, Utilities •1st floor – Electronic lab•2nd ,3rd floors-Offices•Technological Roof –telescopes, satellites antenna, space science experiments

Status: construction

Promising Future (up to 2012)

Basement level•Data Center Infrastructure:

•Redundant 2xUPS 250 KVA•Redundant cooling system•Generator 650KVA

•Computing Facilities:•2000 core - ALICE (AliEn +WLCG)•1000 core – ESA Missions• 500 core - IAF (ISS Analysis Facility)• 250 core - new VO’s (Antares,etc)• 100 core -R&D GRID (testing new GRID technologies, spin-off, etc)

•Storage: •600 TB –ALICE•250TB –ESA Missions•100 TB –IAF•150 TB –new VO’s

Thank you for your attention!

Promising Future New building:

• National Center for Space Science and Technology

•Description:•Basement - Data Center, CleanRoom, Utilities •1st floor – Electronic lab•2nd ,3rd floors-Offices•Technological Roof –telescopes, satellites antenna, space science experiments

Status: construction

Promising Future (up to 2012)

Basement level•Data Center Infrastructure:

•Redundant 2xUPS 250 KVA•Redundant cooling system•Generator 650KVA

•Computing Facilities:•2000 core - ALICE (AliEn +WLCG)•1000 core – ESA Missions• 500 core - IAF (ISS Analysis Facility)• 250 core - new VO’s (Antares,etc)• 100 core -R&D GRID (testing new GRID technologies, spin-off, etc)

•Storage: •600 TB –ALICE•250TB –ESA Missions•100 TB –IAF•150 TB –new VO’s

Thank you for your attention!