Docstoc

sun

Document Sample
sun Powered By Docstoc
					Grid Activity in China


   Sun Gongxing, IHEP, Beijing
The Grid Projects in China
 •   Vega Grid Project ( ICT, MOST)
 •   National High Performance Computing
     Environment funded by 863 program
     (Many participants, MOST)
 • Scientific Data Grid. (CAS, MOST)
 • HEP data grid.(IHEP)
 • Grid infrastructure building (MOE)
Overview of China’s Grid Project
   Build a superserver for major grid sites.
   Build a few of HPC centers.
   Develop system software for the Grid
    system.
   Develop demo applications on the Grid.
   Build 2Gbps interconnection among 100
    university.
    The Vega Project
1. Design goals:
• Design and implement Grid-level software.
• Build a national-scale infrastructure supporting
  HPC.
• Develop   application running on testbed.
2. Grid-oriented superservers—Dawning 4000,
  5000,etc.
3. Service Grid
Stress Distribution of 3800TEU
500hPa Contours in Weather
Forecasting
The Vega Grid Topology

It is Router-centered structure:
The Vega Grid Computing
Protocol
   The Vega OS
Vega OS overview
1, A user-level middleware running on client
  machine
  • Vega GOS constructed by 2 components:
  GOS and gnetd
    - GOS is a daemon running on client machine.
    - gnetd is a daemon on grid server response
      to client commands.
   Grid process and Grid thread
    - Grid process is a unit for managing the whole
      resource of the Grid.
    - Grid thread is a unit for executing computation
      on the Grid.
2, Vega APIs introduction
 Vega GOS APIs

  - GOS APIs for application developers
    grid(): constructs a Grid process on client
  machine.
    gridcon(): grid process connects to Grid
  system.
    gridclose(): close a connected grid.
    ……
• gnetd APIs for service developer on Grid
  servers
  - grid_register(): register a service to
    Grid.
  - grid_unregister(): unregister a service.
  …….
 The Vega Grid Batch System
- A GOS-based application.
- No central control
- shell commands:
  > gsub
  > gstat
  > gkill, ……
- All kinds of resources registered in Grid.
      > Computational resources: cycles, mem,
         disk, networks……
      > software: DB, CAD……
      > services: ftp, http, mail……
Global Batch System in Vega Grid
Travel-An Online Transaction system

  An online transaction system----travel in China
  - GOS-based application

  - 4 different basic services registered in the
     Grid.
    * Wheather forecasting
    * airline ticket booking
    * sight spot ticket booking
    * charging
  - Putting these 4 services into an application
The National HPC Project
   Overview
    - Set up 9 HPC sites across China
    - Interconnected these 9 sites with available
      networks ( Cernet & NSTTet)
    - equipped each site with
      Dawning 2000/3000, Galaxy 3(20Gflops),
      Sunway 1(460Gflops), PC-cluster(8Gflops)
    Overview of the Grid System
    Software

   A function layer on top of node OSs.
   Management of resources, users and
    tasks.
   Authentication and authorization.
   Hardware-enhanced encryption.
   Web –based user interface.
The Grid System Software
User Management
Job Management(submit)
Resource Management
(Find a User)
The Grid System Monitoring
The Grid Utilities
The Grid Applications
   Weather forecasting.
   Petroleum reservoir simulation.
   Bio-information database and applications.
   Numerical wind tunnel simulation.
   Automobile collision simulation.
   Ship structure analysis.
   National scientific databases and applications.
   Digital library.
                           NHPCE
               Node    Node      Node       Node


  Instr.                                              DB
                                                      Info lib.

                      Gigabit IP Network


                           Internet                     Audio


Video
Camera                 DVD


    Tel.                                              Game
           BP,                   PC        Notebook
           Mobile     TV                   HPC
             phone
International Cooperation
   Joint research supported by European
    and China.
   Source sharing
    - Computing power
    - Database
    - Software tools
    - Application software
The Scientific Data Grid Project
   Build on the top of scientific databases.
   Belong to CAS, but open to the world.
   Plan to provide service in 2004-2005.
Scientific Database(SDB)

   SDB funded by CAS since 1986.
   SDB is a collection of multiple disciplinary
    databases, including chemistry, biology,
    geography, astronomy, ecology, etc.
   By 2005, there are
    - 400+ member institutions across China.
    - 300+ databases which is distributed and
      heterogeneous.
    - 10TB+ data volume size.
    Why do SDG?
   Resource level—sharing and development.
    - make data more accessible.
    - data integration
    - and more easy to do: data->information->knowledge.
   Application level—emerging applications
    - can do what we can not do before.
    - cross multiple databases/multiple disciplines.
    - require more resources ( cycles, bandwidth,
      instrument, etc.)
    Tasks

   Build testbed
    - one data center
    - and three sub data center.
   Middleware
    - information service
    - security
    - user interface
   Applications
    - Chemistry/Geo-
    science/astronomy/biology…
        Bio Center
        Cluster 8 nodes     SDG Resources:
        1-2TB                 20 TB
        Beijing               4 PC Clusters
                              CSTNET
1000M

                                 Data Center (CNIC)
                                 Cluster 16 nodes
                                 15TB
1000M
                                      Chemistry Center
                          155M        Cluster 8 nodes
    Geo Center                        1-2TB
    Cluster 8 nodes                   Shanghai
    1-2TB
    Beijing
               SDG Data Center




                                         CA      Portal
                                        Server   Server
Mass Storage    Database      MDS Server
                Application
                Server




                              Supercomputers at CNIC
                                    ~2 TFLOPS
       Grid Middleware
applications                           Application

app-oriented, unified program
                                         GAPI
interface                                            xMDS
coordinated access to multiple data
                                          DRB
resources

universal access interface to single       UAI
data resource                                        GSI
local data management system, could    Local DBMS
be DBMS or file system

databases
Projects
   CAS
    The tenth five-year program (2001—
    2005) (37MRMB).
   863 program (MOST)
    A special program (proposed).
Collaboration
   PRAGMA
   APGrid
   SDSC
   KISTI
   ASCC
   Texas A&M Univ.
HEP Data Grid
   Currently funded by IHEP, CAS.
   Build a BES VO in IHEP campus network
    environment for BES experiment.
   Probably Add to CMS and Atlas Vos in
    the future for CMS and Atlas
    collaborations.
   ARGO-YBJ VO for ARGO-YBJ experiment.
The HEP Grid Testbed
Topology
              GIIS
                      Two 6 CPUs PC-FARM



     GRIS            GRIS




     Condor                 PBS
Current Main Tasks
   Large Scale Linux Configuration and
    maintenance—LCFG.
   Authentication/authority—GSI.
   Resource Information Management—MDS.
   Grid Resource Allocation Management—
    GRAM.
   Data Replication Management—GDMP.
   HEP analysis software gridifying—Appln.
    The Future Work

   Build IHEP campus Grid computing
    environment sharing about ten PC-Farms.
   Promote collaboration with National Grid
    Projects and international Grid communities
   Make effort to propose Regional Center 2
    project in China.

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:6
posted:7/8/2011
language:English
pages:44