Grid Computing: An Overview

1
Grid Computing An Overview

• Vijayanand Bharadwaj

2
Introduction

• Motivation
• A new class of application are emerging.
• Deal with Large Data Sets
• Distributed sources of input
• Monitoring and analysis, pattern recognition
• Real-time response
• Very often data sets are disjoint
• Asynchronous input
• Resource-intensive
• Reliable (Fault-Tolerant)

3
Introduction

• Motivation (continued)
• Examples
• Public-health surveillance
• Monitoring of outbreaks by examining
  non-identified patients
• Bio-terrorism
• Financial market analysis
• Infrastructure security surveillance
• Electric power grid security, power outages,
  blackouts
• Cyber-security surveillance
• Intrusion detection algorithms
• National security surveillance
• Monitoring communication, (cyber and voice)
• Entry and exit of passengers at national ports

4
Introduction

• Motivation (continued)
• Need infrastructure (hardware software) and
  mechanisms which can handle such applications,
  Especially
• Support demands for resources, computing speed,
  memory etc.
• Reliable (Fault-Tolerant)
• Communicate using existing networks and protocols
  ( such as the public Internet) as inputs are
  distributed
• Secure

5
Introduction

• Motivation (continued)
• Enter the GRID !!!
• Use of a Grid as a computing platform.
• Just as the Internet provided a substrate for
  running n-tier applications, the GRID can be used
  to run resource-intensive applications in a
  secure, fault-tolerant manner.

6
Overview

• What is a GRID?
• Grid and the Internet
• GRID v/s Cluster
• Grid Computing Infrastructure
• Standards OGSA WSRF
• The Globus toolkit

7
What is a GRID ?

• A type of parallel and distributed system that
  enables the sharing, selection, and aggregation
  of geographically distributed "autonomous"
  resources dynamically at runtime depending on
  their availability, capability, performance,
  cost, and users' quality-of-service
  requirements. 1
• geographically distributed resources
• selected based on capabilities and availability
• dynamically at run-time

8
What is a GRID ?

• Figure from 2

9
GRID Internet Where is the connection ?

• GRID can have its own dedicated network to
  connect different points
• High Speed, Secure, High Available etc.
• But penetration depends on cost !
• The Internet provides the substrate for a GRID.
• The connectivity (Internet Architecture)
• The Protocols and Standards for a GRID Standard
  (HTTP, Web Services)
• Ubiquitous as the Internet (advantage)
• Must deal with the Internets problems,
  availability, security etc. ( disadvantage)

10
GRID versus Cluster

• Cluster
• Homogeneous
• Single location,
• Dedicated Ethernet connection on the same board
• Share the data bus,
• Very low latency,
• Capacity of cluster known ( i.e. upper limit can
  be a bottleneck ),
• Availability guaranteed,
• Must own and administer one.

• Grid
• Heterogeneous systems
• Distributed
• Connected over high speed dedicated links or
  public existing networks (Globus)
• Latency varies
• No known upper limit (an advantage
• Availability not guaranteed,
• Need not own, can subscribe and submit jobs

11
GRID versus Cluster

• A Grid connects Clusters ! ! ! !
• Multiple Clusters Connected by a Grid
• Each Cluster autonomous
• Provides Computing Power, Memory
• Harnessed by user at run-time
• Jobs are divided and submitted, and then results
  are aggregated

12
Grid Standards

• How can multiple autonomous heterogeneous
  computing resources be connected, and work
  together ?
• How can their resources be discovered, selected
  and harnessed at run-time?
• How can users communicate with these resources ?
• How do they know the jobs are completed ?
• How do they authenticate themselves
• Many other questions..
• SO NEED STANDARDS !

13
Grid Standards

• Requirements for a Grid
• 1. Resource Discovery Management Service
  (Information Services)
• 2. Job Management Service
• 3. Security Service
• 4. Data Transfer Service

14
Grid Standards 5

• OGSA (Open Grid Services Architecture)
• Developed by the Global Grid Forum (GGF) 7
• Specifies a Standardized Open Architecture for
  all the services required for a Grid.
• Grid Services are defined by OGSA.
• Grid Service
• A program which completes a task, computational
  capability, memory, storage etc offered by an
  entity. These can be uniquely specified,
  discovered and invoked remotely by clients. Based
  on Web Services.
• OGSA defines what Grid Services are,
  capabilities, technologies

15
Grid Standards 5

• BUT YOU NEED MORE !
• How does the Grid Work ?
• You have the Internet to connect the resources.
• But how can they interoperate ?
• You need some Middleware !
• Enter WEB SERVICES !

16
Grid Standards 5

• WSRF -- Web Services Resource Framework
• Defined by OASIS 8
• Defines Stateful Web Services
• What are Web Services
• Ability to discover, invoke programs remotely
  using Internet related protocols such as HTTP and
  concepts such as Uniform Resource Indicator (URI)
  ( Similar to URL)
• Similar in concept only to CORBA, RMI, EJB notion
  but many differences in the details
• Use of XML based means to transfer data
• Has a suite of Protocols and Standards
• Simple Object Access Protocol
• Web Services Description Language and many others
  !

17
Grid Standards 5

• What are Web Services ?
• Ability to discover, invoke programs remotely
  using Internet related protocols such as HTTP and
  concepts such as Uniform Resource Indicator (URI)
  ( Similar to URL)
• Mechanisms with which services can be uniquely
  specified, discovered and invoked.
• Similar in concept only to CORBA, RMI, EJB notion
  but many differences in the details
• Platform Independent, language independent (use
  XML based languages such as WSDL, SOAP)
• Use of XML based means to transfer data
• Work over existing networks and protocols
• Has a suite of Protocols and Standards
• Simple Object Access Protocol
• Web Services Description Language and many others
  !

18
Grid Standards 5

• Web Services An example

19
Grid Standards 5

• Web Services (contintued)
• Loosely coupled C-S model unlike CORBA and EJB
• Overhead is a disadvantage
• Figure Web Service Interaction. From 9
  (Definition from Older Version of Globus Toolkit
  (GT3) Manual Good for some Pictures !)

20
Grid Standards 5

• Web Services (contintued)
• Figure Web Service Architecture Service
  Invocation. 9 (Definition from Older Version of
  Globus Toolkit (GT3) Manual Good for some
  Pictures !)

21
Grid Standards 5

• But Web Services are Stateless !
• Do not maintain state of the current service, and
  respond to the requestor
• Similar to the HTTP Request-Response paradigm (
  In fact it is due to HTTPs statelessness as Web
  Services use HTTP)
• Grid applications need to maintain state !
• Computationally Intensive !
• So WSRF is the Stateful standard of the Web
  Services and Grid is based on WSRF

22
Grid Standards 5

• OGSA WSRF

23
The Globus Toolkit

• Software suite which creates a grid
  infrastructure over distributed hosts on which
  applications can be run.
• An Implementation of the OGSA
• Created by the The Globus Alliance 3
• (Argonne Natl. labs, Univ. of Chicago, USC, etc)
• Globus Alliance a member of the GGF
• Open source 3 Commitment to Open Source The
  Globus Alliance is committed to maintaining a
  liberal, open source license. The Globus Toolkit
  Public License (GTPL) allows software to be used
  by anyone and for any purpose, without
  restriction. We believe that this is the best way
  to ensure that Grid technologies gain wide spread
  acceptance and benefit from a large developer
  community.

24
The Globus Toolkit 5
25
The Globus Toolkit

• Main Components of a Grid 4
• Security- Ensure that only authorized users can
  access and use Grid resources.
• Data management- Data must be transported,
  cleansed, parceled, and processed.
• Resource management - The grid must know what
  resources are available for different tasks.
• Information services- Users and applications
  must be able to query the grid efficiently.

26
Main Components of Globus 3
GLOBUS ToolKit
Runtime Components Java WS Core C WS Core XIO
Security GSI
Information Services MDS (GRIS/GISS)
Data Management GridFTP RFT RLS
Resource Management GRAM GASS
27
Globus Components

• The main components are 3
• Runtime Components
• Java WS Core, C WS Core, XIO
• Resource Management
• GRAM Grid Resource Allocation Manager
• GASS Global Access to Secondary Storage
• Data Management
• GridFTP Protocol for high-speed data transfer
• RFT
• RLS

28
Globus Components

• Information Services
• MDS Monitoring and Discovery Service
• GRIS (Grid Resource Information Service)
• GIIS (Grid Index Information Service)
• Security
• GSI
• Grid Security Infrastructure

29
Globus Architecture (older Version GT3)

• Figure Globus Architecture. From 6 (Note This
  is older version GT3, but similar concept, good
  picture)

30
GRAM (older Version GT3)

• Figure Working of GRAM protocol. From 6 (Note
  This is older version GT3, but similar concept,
  good picture)

31
Globus Application Software

• Grid Application Software
• Structure and content is very similar to Web
  Services applications.
• Uses same protocols, languages and paradigms.
• Creating Grid Applications is very similar to
  Creating Web Services Applications ! !
  (advantage)

32
References

• 1 Grid definition, http//www.cs.mu.oz.au/ra
  j/GridInfoware/gridfaq.html
• 2 Fundamentals of Grid Computing,
  REDP-3613-00 IBM Redpaper, published
  November-12-2002, http//www.redbooks.ibm.com/red
  papers/pdfs/redp3613.pdf
• 3 The Globus Alliance http//www.globus.org
• 4 IBM developerWorks staff, Start here to
  learn about Grid Computing, August 2003
• http//www-106.ibm.com/developerworks/grid/li
  brary/gr-starthere.html
• 5 The GTK4 Programmers tutorial
  http//gdp.globus.org/gt4-tutorial/
• 6 Introduction to Grid Computing with Globus,
  SG24-6895-01Redbook, published October-1-2003,
  last updated October-6-2003 http//www.redbooks.ib
  m.com/redbooks/pdfs/sg246895.pdf

33
References

• 7 Global Grid Forum http//www.ggf.org
• 8 OASIS http//www.oasis-open.org/
• 9 The GTK3 Programmers tutorial
  http//gdp.globus.org/gt3-tutorial/

34
Thank You

• Questions, Comments, Criticisms
• welcome at
• vijay_at_csee.wvu.edu