Rio GRID Initiatives : T2HEPGRID - PowerPoint PPT Presentation

1 / 18
About This Presentation
Title:

Rio GRID Initiatives : T2HEPGRID

Description:

1985 - ACP2 ( hardware: board design , software: OS ) ... Link provided by REDE RIO at 2 Mbps. 2003- Start to buy the machines and build the group ... – PowerPoint PPT presentation

Number of Views:44
Avg rating:3.0/5.0
Slides: 19
Provided by: alberto8
Category:

less

Transcript and Presenter's Notes

Title: Rio GRID Initiatives : T2HEPGRID


1
Rio GRID Initiatives T2-HEPGRID
Andre Sznajder UERJ(Brazil) http
//www.hepgrid.uerj.br
2
HEP Computing in Rio
  • 1982 Begining of Experimental HEP activities
  • 1983 - ACP1 ( paralelization software CPS )
  • 1985 - ACP2 ( hardware board design , software
    OS )
  • 1995 CHEP 95 in Rio ( first international VC in
    Brazil )
  • 1997 Client/Server project with IBM(Grid
    Precursor )
  • Client/Server Upgrade GRID

3
HEPGRID Historic
  • 2001- Presentation of a New Project to FINEP
  • Aproved 500 machines But...
    Dollar increase 3x
  • could buy only 100
    machines.
  • 2002- The first part of the financial support
    delivered.
  • Bureaucracy delays our projects
    about 2 years
  • Need to create infrastructure.
  • Link provided by REDE RIO at 2
    Mbps
  • 2003- Start to buy the machines and build the
    group
  • CALTECH group has been very
    helpful Many Thanks !
  • 2004- Hepgrid and Digital Divide Workshop at
    UERJ,
  • See http//www.lishep.uerj.br/
  • December 20th. inauguration of UERJ HEPGRID
    cluster

4
HEPGRID Cluster at UERJ
  • 1 Frontend Pentium Xeon Dual CPU 2.7 GHz ,
  • 4 GB RAM , 2x 36GB
    SCSI HD , 2x 1Gbit NIC
  • 82 Nodes Pentium Xeon Dual CPU 2.7 GHz , 1 GB
    RAM ,
  • 40 GB HD , 1Gbit NIC
  • 3 TB Raid disks ( 2 NAS File Servers )
  • Five 24 Gbit port switches for internal
    connectivity
  • Operational System Rocks 3.3.0 ( based on
    RedHat EL3.0 )
  • Development cluster for testing 1 frontend , 3
    nodes
  • GRID software OSG 0.2.1

5
HEPGRID Cluster
RAID DISKS
NODES
NAS
FRONTEND
NOBREAKS
6
Cluster Network Topology
7
Cluster Monitoring (GANGLIA)
8
Cluster Monitoring (GANGLIA)
9
Cluster Monitoring (MONALISA)
10
External Network Connectivity
CLARA 100Mbps
11
External Network Connectivity
12
Rio 2005 CMS GRID Meeting
  • Main outcomes
  • Decision to connect HEPGRID to Fermilab T1
  • Grid3/OSG installation certification
  • Start CMS Monte Carlo production
  • RNP decided to provide cluster connectivity
  • (Currently 100Mbps , but will be upgraded to
    1Gbps exclusive
  • after September 2005)

13
Rio CMS GRID Meeting
14
National Research and Education Network (RNP)
  • Maintained by the Brazilian government
  • Provides national (inter-state) and international
    connectivity for more than 200 universities and
    research centers
  • collaboration links to other similar networks
    internationally (Internet2, GÉANT, APAN,
    RedCLARA)
  • supports the development of advanced networking
    and its applications

15
RNP Network Topology
16
SC2004 Network Traffic Between UERJ CERN
Thanks to RNP and Caltech group !
UERJ to CERN
CERN to UERJ
17
HEPGRID Team
18
Perspectives
  • Install SAMGRID and begin P17 reprocessing MC
    production for D0
  • Install dCache for CMS digi/reconstruction
  • Storage upgrade to 200TB
  • Cluster upgrade ( rack mounted ? )
Write a Comment
User Comments (0)
About PowerShow.com