Grid computing [2005]

GrId
ComputIn
g
GSI640
Raúl Soto
Joel Maysonet
Lucette Sánchez
Fernando González

What is Grid Computing?
 A Grid Computing system is a
collection of distributed computing
resource available over a local or
wide area network, that appears
to an end user or application as
one large virtual computing
system

 Is an approach to distributed
computing that spans not only
locations but also organizations,
machine architectures, and
software boundaries

Characteristics
 Allows the integrated, collaborative use of computers, networks,
databases, and scientific instruments owned and managed by
multiple organizations

 Facilitates the solution of computational problems
– large-scale
– complex
– multi-institutional
– multidisciplinary
– Large data storage and/or computational requirements

 Grid computing started out as the simultaneous application of the
resources of many networked computers to a single (usually
scientific) problem.
– E.G. : SETI @ Home, Human Proteome project, Anthrax research,
Smallpox project, Cancer research project, etc.

Characteristics
 For many years, computational grids have been used to solve
large-scale problems in science and engineering

 Currently used in the following fields :
– Medical research : protein folding, cancer drug development
– Astronomy : SETI data analysis
– Mathematical / Statistical problems
– Climate models

 Grid computing is beginning to enter the commercial world
– Financial analysis
– Forecasting
– Enterprise Grids

Characteristics
 Involves sharing of heterogeneous resources:
– Hardware platforms
– Hardware / software architectures
– Computer languages
– Different places
– Different administrative domains

Main Classifications
 Grid Types:
– Computational Grids : computers set aside resources
allocated to number-crunch data or provide coveratge for
CPU-intensive workloads
– Data Grids : share data resources and storage capacity,
unified interface for all data repositories in an organization,
through which data can be queried, managed, and secured
– Scavenging Grids : used to locate and exploit machine cycles
on idle servers and desktops for use in resource-intensive
tasks
 Internal vs External Grids
– External grids : usually geographically-distributed, non-profit
research efforts
– Internal grids : large commercial enterprise with complex
problems who aim to fully exploit their unused internal
computing power

How does it work?
System users
Scientists and engineers Grid Operating System
Using computation to The software which
coordinates the interplay Software
Accomplish Lab missions
of computers, networks, Software applications
Intelligent Interface and storage and components for
A knowledge-based computational problems
environment that offers users
guidance on complex Hardware
computing tasks Heterogeneous collection
of high-performance
computer hardware and
software resources

Networking
The hardware and
software that permits
communication among
distributed users and
Middleware computer resources
Software tools
that enable Mass Storage
interaction A collection of devices
among users, and software that allows
applications, and temporary and long-term
system archival storage of
resources information

Grid Computing vs the Internet

 The Internet is about getting computers to talk
together
 Grid computing is about getting computers to work
together

• The Internet is a network of
communication

• Grid computing is a network of
computation: provides tools and
protocols for resource sharing of a
variety of IT resources

Grid Computing vs.
Clusters / Distributed Computing

 Clusters and Distributed computing systems require
– physical proximity
– Operational homogeneity
 Grids are geographically distributed, heterogeneous

 Clusters and Distributed computing systems are
based on the Data Center / Computer Center
computers
 Grids include end-user computers as well

Analogy: Electric Power Grid
 Electric Power:
– Household electrical devices simply plug to an electric outlet
– Use only the electric power you need
– Pay only for the electric power you used

 Computers – typical :
– You pay for a computer with certain computing power (CPU flops)
– If you use less computing power than what your computer
provides, you still pay for all of it
– If you use more computing power than what your computer
provides, you have to buy a better computer

 Computers – Grid :
– Multiple computers (including yours) share computing resources
and make up a virtual computer
– You only use the computing power you need

Grid Computing
 Computing power cost
– 1980: $100,000 per megaflop
– 2000: $1 per megaflop

 Computing power evolution
– 1986: US National Science Foundation resources were five
(5) Cray XM-P supercomputers
– 2000: that’s the equivalent computing power of ONE (1)
Nintendo 64 console

Applications
 Distributed supercomputing /
computational science

 High – capacity / throughput
computing: large-scale simulation,
chip design, and parameter studies

 Content sharing: for example,
sharing digital content among peers

 Remote software access / renting

Applications
 Data-intensive computing: drug
design, particle physics, stock
prediction

 On-demand, real-time computing:
medical instrumentation, mission
critical initiatives

 Collaborative computing (e-science,
e-engineering), collaborative
design, data exploration

Benefits

 Permits sharing of resources throughout an
organization, or among organizations

 Make effective use of underused computing resources

 Provide access to remote databases and software

 Reduce significantly the number of servers needed (25-75%)

 Allow on-demand aggregation of resources at multiple sites

 Reduce execution time for large-scale data processing
applications

Benefits
 Provide load sharing across a
set of platforms

 Provide fault tolerance

 Take advantage of time-zone and random diversity (in peak
hours, users can access resources in off-peak zones)

 Provide flexibility to meet unforseen emergency demands: can
rent external resources for a required period instead of buying
additional capacity

 Virtual data centers

Disadvantages
 Proprietary approaches
– Leading vendors (HP, Sun, IBM, MS, Oracle, etc.)
have proprietary, incompatible approaches,
– which defeats the purpose of Grid computing

 Business case not always easy to sell to upper management
– Need to present a business case based on economics, efficiency, not
technical details

 Vendors need to show how their software effectively manages a grid
environment

 Security
– Confidentiality, Integrity, Access to resources, data

 Performance monitoring
– Zoning mechanisms to ensure applications competing for resources do not
affect each other

Evolution of Grids
Performance
and QoS

Global Grid

Partner Grids

Enterprise
Cluster / Grid
Local
Data
SMPs or Local Cluster Grids
Super Computing
computers
Personal
Devices

Early Stage Second Stage Third Stage
1990s Early Mid Late 2000s
2000s 2000s

[source: Minoli, Daniel, A Networking Approach to Grid Computing, 2005]

Enterprise Grids Example : AstraZeneca PLC Grids

 Data Grid
– Connects R&D databases from sites in UK, Sweden, and USA
– Significant savings in finding information
– Efficiency gains due to shortening the time R&D or design staff
needs to find information
– Large investment in broadband links to connect data centers in
different countries

 Enterprise Grid
– Processing power shared between R&D sites in UK, Sweden, and
USA
– Efficiency due to processing power sharing, plus access to data
– Savings on R&D time, time to market
– Permits more efficient collaboration between sites
– Significant investment in security, and in high-performance
broadband links

Some Examples of Grids
Grid Name Sponsor Purpose
BlueGrid IBM IBM computation R&D

DISCOM Sandia National Labs Defense research

DOE Science Grid DOE Office of Science Scientific research

European Union DataGrid European Union Scientific research

EuroGrid GRIP European Union Computation R&D

Globus Project DARPA, NASA, Msoft, others Grid tech research

GridLab European Union Grid tech research

Grid Research Integration National Science Foundation Grid middleware developmt

Intern. Data Grid Lab National Science International large scale
Foundation grid tech research
Information Power Grid NASA Aerospace research

Earthquake Eng. Simulations National Science Foundation Earthquake engineering

Particle Physics Data Grid DOE Science High-energy physics research

TeraGrid National Science Link major US
Foundation universities
UK Grid Support Center UK eScience Grid projects in UK

TeraGrid

(www.teragrid.com)
 Completed in September 2004

 Massively parallel supercomputer clusters

 40 teraflops of computing power

 2 petabytes of rotating storage

 Connected network of US supercomputing
centers (currently 8, and growing)
 Each of the four original sites operates a Linux
cluster, interconnected by means of a 10-30
Gigabit/sec dedicated optical network

Pauá Grid
Project - Brasil

 Partnership with HP
 Links various academic sites
across the nation
 7 sites, 200 nodes, 1900 miles
 Bioinformatics, data mining,
security applications
 Attracts projects from other
knowledge areas that need high
performance computing

Grid Computing Applications
 SETI@home

 Human Proteome project

 Anthrax research project

 Smallpox project

 Cancer research project

 SciRun environment

SETI@home
 LARGEST GRID COMPUTING
PROJECT IN HISTORY
 Runs on over 50,000 PCs
 Generates approximately
1,000 CPU years / day
 Approx 500,000 CPU years so far
 Averages 40 Tflops/second
 Distributes data sets from the Arecibo Radiotelescope
 Performs sophisticated data and signal processing analysis
 Can run as a screen saver or as a
low priority process

http://setiathome.ssl.berkeley.edu

Human Proteome Folding Project
 Find functions for all the proteins
encoded in the Human Genome
 When human protein structures are
known, scientists can use them to
research disease treatments and cures
 Only a fraction of 30,000 human
http://www.grid.org/projects/hpf/ proteins have known structures and
functions
– Examining the entire human
genome could require up to 1
million years of computational time
on a Pentium IV.
– Using a commercial 1000-node
cluster would require 50 years and,
while faster, would still be
impractical.
– Can run as a screen saver or as a
low priority process

Anthrax Research Project
 As of February 14, 2002, the screening phase of the Anthrax
Research Project has been completed. In 4 weeks it achieved
what previously took years.

 This project's goal was to accelerate what is usually a time-
consuming step in the lengthy drug discovery process.

 The project entailed presenting a key protein component of
anthrax into the general rotation of the United Devices Member
Community's current virtual screening project, which works with
the Grid MP platform over the Internet.

 This allowed UD Members to lend their computers in the
screening of 3.57 billion molecules for suitability as a treatment
for advanced-stage Anthrax.

Anthrax Research Project
 Screening is only one step in a long drug discovery process that
ultimately must move from the computational realm into the
actual laboratory.

 The project used a 5-time redundancy rate for each molecule to
ensure a high level of accuracy and quality.

 Preliminary indications are that we have narrowed the original
pool of 3.57 billion molecules down considerably, having
identified over 300,000 crude unique hits in the course of the
project.

 This significantly reduces the next phase of the discovery
process, in which the ranked hits will be further refined and
analyzed, accelerating the overall time to availability of a
treatment.
http://www.grid.org/projects/anthrax/

Cancer Research Project
 Processes molecular research being conducted by the
Department of Chemistry at the University of Oxford in England
and the National Foundation for Cancer Research.

 To participate, users download a very small, no cost, non-
invasive software program that works like a screensaver: it runs
when your computer isn't being used, and processes research
until you need your machine. Your computer never leaves your
desk, and the project never interrupts your usual PC use.

 The research centers on proteins that have been determined to
be a possible target for cancer therapy. Through a process
called "virtual screening", special analysis software will identify
molecules that interact with these proteins, and will determine
which of the molecular candidates has a high likelihood of being
developed into a drug.

Cancer Research Project
 The process is similar to finding the right key to open a special
lock — by looking at millions upon millions of molecular keys.

 It allows computers to screen molecules that may be developed
into drugs to fight cancer. Each individual computer analyzes a
few molecules and then sends the results back over the Internet
for further research.

 This project is anticipated to be the largest computational
chemistry project ever undertaken and represents a genuine
hope to find a better way to fight cancer.

 The computational power to perform research of this scale is
only available through the generosity of participants.

http://www.grid.org/projects/cancer/

Smallpox Research Project
 Smallpox was eliminated from the world in 1977
 Stocks of the variola virus still exist, potential terrorist use
 Vaccination ended in 1972, so an outbreak would kill millions
 There is a possible molecular target whose blockade would
prevent the ravages of an infection.
 The Smallpox Research Grid Project involves screening 35
million potential drug molecules against several protein targets
— one of the largest computational chemistry project ever
undertaken.
 This will involve the use of the United Devices Grid MP Global,
which we have successfully used in the past towards cancer and
anthrax research.
 The project can harness millions of computers belonging to
people in over two hundred countries, all of whom will benefit
from protection against smallpox.
 Can run as a screen saver or as a low priority process

http://www.grid.org/projects/smallpox/about.htm

SCIRun
 Interactive Scientific Programming
 Interactively composes large-scale scientific computations
through visual dataflow programming
 Integrated with visualization packages
 Supports interactive steering during the design, computation,
and visualization phases of a simulation

http://www.cs.utah.edu/sci/scirun

Key Vendors

 Parabon Computation www.parabon.com
 DataSynapse www.datasynapse.com
 IBM Grid Computing www.ibm.com/grid
 Sun Microsystems Grid Computing www.sun.com/grid
 Oracle Corp. www.oracle.com/grid
 HP Grid Computing
www.hp.com/techservers/grid/index.html
 United Devices www.ud.com
 1st Port for Grid Computing www.1stport.co.uk

References
 Computational Intermediation and the Evolution of Computation as a
Commodity, Applied Economics, June 2004
www.business/duq.edu/faculty/davies/research/economicsofcomputation.pdf

 The Grid: Blueprint for a New Computing Infrastructure
www.mkp.com/grids

 Grid Computing: Making the Global Infrastructure a Reality
www.grid2002.org

 The Grid : Core Technologies
http://coregridtechnologies.org

 CERN: The Grid Café – What is a Grid?
http://gridcafe.web.cern.ch/gridcafe/whatisagrid/whatis.html

 Grid Computing: A Brief Technology Analysis
www.ctonet.org/documents/gridcomputing_analysis.pdf

 IBM: What is Grid Computing
http://www-1.ibm.com/grid/about_grid/what_is.shtml

References
 IBM Grid Computing Benefits
http://www-1.ibm.com/grid/about_grid/benefits.shtml

 Sun Microsystems: What is Grid?
http://www.sun.com/executives/iforce/integratedsolutions/gridsolutions/index.html

 HP Grid Computing
http://www.hp.com/techservers/grid/index.html

 Oracle Grid Index Report
http://www.oracle.com/global/eu/pressroom/nagridreport.pdf

 Juhasz, Zoltan, et al, Distributed and Parallel Systems, Cluster and Grid
Computing , Springer Science and Business Media, 2005 [ISBN 0-387-23094-7]

 Minoli, Daniel, A Networking Approach to Grid Computing, John Wiley and Sons,
2005 [ISBN 0-471-68756-1]

 Wikipedia: Grid Computing GSI640
http://en.wikipedia.org/wiki/Grid_computing

Grid computing [2005]

Recommended

More Related Content

What's hot (20)

Similar to Grid computing [2005] (20)

More from Raul Soto (19)

Recently uploaded (20)

Grid computing [2005]