Grid Computing Adoption in Research and Industry

1
Grid Computing Adoption in Research and Industry

• University of Tehran
• College of Engineering
• Department of Electrical Computer Eng.
• Parallel Processing
• Fall 2007

2
Outline

• Grid Evolution Review
• Grid Advantages
• Global Grid
• Components of Grid
• DRM
• Globus
• Access Portal
• Grid Examples

3
Grid Evolution Review

• Three domains
• Cluster Grid
• Campus Grid
• Global Grid

4
Grid Advantages

• Key advantages Grid can provide to research and
  industry
• Access (to computing, data, experiments,
  instruments, sensors, etc.)
• Virtualization
• On Demand
• Sharing (jointly work on one complex task.)
• Failover (migrate)
• Heterogeneity
• Utilization
• Benefits to managers
• Increased agility
• Reduced risk
• Enabled innovation

5
Global Grid

• Why move to Global Grid?
• Further Increasing the throughput by maximizing
  resource utilization.
• Increasing the range of complementary hardware
  available.
• Providing a (virtual) supercomputer for Grand
  Challenge applications.
• Main software components of Global grid (as a
  whole)
• User interface a transparent, secure interface,
  usually a Web-portal.
• Broker automating job scheduling based upon the
  users policies.
• Security, data-management, job-management, and
  resource discovery.
• Resource guarantees and accounting.

6
Components of Grid

• Core components of a Grid
• Access Portal to plug into a Portal server
• e.g. Grid Engine Portal
• Infrastructure Toolkit to glue all components
• e.g. Globus Toolkit
• Distributed Resource Management
• e.g. Sun Grid Engine, Sun Grid Engine Enterprise
  Edition

7
DRM

• Distributed Resource Manager
• Gathers resource requests and maps compute jobs
  to the least-loaded and best suited system in the
  network.
• Necessary where the underlying compute resources
  are heterogeneous
• operating platform
• processor architecture
• memory
• Local DRM provides a virtualization of these
  resources, usually by means of the queue concept.
  
• Queue represents the underlying compute resource
  to which jobs are submitted.

8
Globus

• Globus is the most famous infrastructure lies on
  top of DRMs and manages them. Also it provides a
  unique interface (Grid) to access portals and
  users.
• Has three components
• Globus Security Infrastructure (GSI) provides
  the underlying security for the Globus
  components.
• GSI is based upon Public Key encryption.
• Provides the mutual authentication between the
  hosts involved.
• Globus Resource Allocation Manager (GRAM)
  provides the ability to submit and control jobs.
• GRAM includes the RSL Resource Specification
  Language in which users can describe their job
  requirements.
• Monitoring and Discovery Services (MDS) provides
  the ability to discover the available resources
  on the grid.
• MDS implements a hierarchical structure of LDAP
  databases.

9
Access Portal

• Divides into two parts
• Web server and/or container.
• e.g. Sun ONE Portal Server, Tomcat/Apache,
  uPortal.
• Collection of Java servlettes, Web services
  components, Java beans, etc.
• make up the interface between the user and the
  Globus Toolkit and that run within the Server.
• e.g. The Grid Portal Development Kit

10
Grid Example (1)

• Houston University Campus Grid
• A Campus Grid
• Researchers need significant computational
  resources.
• e.g. chemistry, geophysics, mechanical
  engineering, computer science and mathematics
• For example hardware at the High Performance
  Computing Center (HPCC) a cluster of Sun Fire
  6800 and 880 platforms connected via Myrinet
  running SGE as DRM.
• EZ-Grid system is on top of Globus as middleware.

11
Grid Example (2)

• High Performance Computing Virtual Laboratory
  (HPCVL)
• A Global Grid
• Formed by four Ontario universities (Carleton
  University, Queens University, Royal Military
  College, and the University of Ottawa)
• High performance researches
• e.g. in population health, photonics, psychology,
  economics, nuclear physics, civil engineering,
  nanomaterials, applied mathematics,
  bioinformatics, and massive data handling.
• 1024-bit encrypted secure Web-based

12
Grid Example (3)

• National Research Council of Canadas Canadian
  Bioinformatics Resource (NRC-CBR)
• A Global Grid
• Collaborating institutes, universities, and
  individuals across Canada
• For research in distributed Bioinformatics
• A computationally massive resource as well as a
  massive database

13
Grid Example (4)

• White Rose Grid (WRG)
• Global Grid
• Based in Yorkshire, UK, comprising three
  universities The universities of Leeds, York,
  and Sheffield.
• There are four significant compute resources
  cluster grids.
• The White Rose Grid is heterogeneous in terms of
  underlying hardware and operating platform.
• command line and GUI of GEEE is the main access
  point to each node for local users

14
Grid Example (5)

• Progress Polish Research on Grid Environment on
  Sun Servers
• Global Grid
• It involves two academic sites in Poland Cracow
  and Poznan
• Two cluster grids Sun Fire 6800 (24 CPUs) in
  Cyfronet Cracow and two Sun Fire 6800 (24 CPUs)
  in Poznan for computing resources for the Polish
  scientific community
• Globus Toolkit 2.0 (and 2.2) have been
  implemented to provide the middleware
  functionality
• System Modules
• Portal Environment
• Grid Service Provider
• Grid Infrastructure
• Data Management System

15
References Links

• Main reference
• A. Abbas, Grid Computing A Practical Guide to
  technology and Applications, Charles River Media,
  Inc. 2004, Chapter 16.
• Links
• Sun Grid Engine Web site, http//www.sun.com/grid
  
• Grid Engine open source project,
  http//gridengine.sunsource.net/
• Globus Web site, http//www.globus.org
• Sun ONE Studio, http//wwws.sun.com/software/sunde
  v/
• Apache Tomcal server, http//jakarta.apache.org/to
  mcat/
• GlobeXplorer Web site, http//www.globexplorer.com
  
• HPCVL Web site, http//www.hpcvl.org
• Canada NRC-CBR, http//cbr-rbc.nrc-cnrc.gc.ca/
• White Rose Grid, http//www.informatics.leeds.ac.u
  k/pages/05_facilities/06_grid.htm
• PROGRESS project home page, http//progress.psnc.p
  l/