Grid Computing: The Way of the Future?Mike Beck, Christopher Koch, Patrick Gerber, Adam MeyerCSC 272 – Spring 2012

Grid Computing…

• Imagine…• Millions of computers• Many people• Worldwide

• Connect them all to a single supercomputer• Powerful calculations

and analysis• As easy as plugging in

an outlet

History of Grid Computing

• Idea from developers of the Multics O.S.• 1965• Access computing resources

like water, gas, electricity• Pay accordingly

• Metacomputing• 1980• Larry Smarr• Interconnecting

supercomputer centers to achieve super processing resources

• Ancestor to grid computing

Source: http://farm1.staticflickr.com/37/88999826_b0c4fe137b.jpg

History of Grid Computing

• Information Wide Area Year (I-WAY)• Infrastructure built upon

metacomputing concept• Strongly influenced following

computing activities• Ian Foster

• Foster-Kesselman duo• 1997 paper that linked

metacomputing to Global Toolkit

• Toolkit for grid software development today

• Workshop titled “Building a Computational Grid” Ian Foster

Source: http://lifeboat.com/board/ian.foster.jpg

History of Grid Computing

• Foster-Kesselman duo (continued)• 1998 book “The Grid:

Blueprint for a New Computing Infrastructure”

• Widely considered “Grid Bible”• Thus, Foster and

Kesselman considered “fathers” of grid computing The “Grid Bible”

Source: http://www.ieuc.org/graphics/bibliography-images/3674893056Foster.gif

What is Grid Computing?

• Accepted definition• system that coordinates resources

that are not subject to centralized control

• using standard, open, general-purpose interfaces and protocols

• to deliver nontrivial qualities of service.

• Unrestricted technology• Any type of computing

resources into a single environment• To perform demanding tasks

that would normally require large-scale computers

• Effectively using resources that would otherwise remain idle.

What is Grid Computing?

• Can be dedicated to a specific application, but it is more common that a single grid will be used for a variety of purposes.

• Distributed system of nodes with non-interactive workloads

• Special type of supercomputing that relies on computers connected to a private or public network

• Nodes tend to be loosely coupled, heterogeneous, and geographically dispersed

• Typically used within scientific applications and researching

Design Considerations

• 3 major types of grids• Computational Grid

• high-performance computing in processor-demanding tasks.

• Data Grid• Made up of a large number of

smaller storage devices• Network Grid

• High-performance network services, providing facilities to speed up network communication

• All grids follow same simple structure• A bottom-up implementation makes

sure that everything in a single layer is working properly before moving up to the next layer

Middleware

• Provides services to other software applications

• Usually runs as a layer over the operating system to provide a common interface

• Developers don’t have to worry about non-core functionality

• Organize and integrates the resources of the grid

• Automates all of the interactions between nodes

• Not a single program

• Agents

• Brokers

• Deals

• Housekeeping agents• All automatic at a fraction of the time• Common protocols for communication,

authentication, and connectivity

Design Hazards

• The features that make grid computing unique and advantageous make it dangerous

• Security is a high risk - solely by trust

• Nodes must trust user will not abuse their access

• Interfering with concurrent operations, retrieving private/sensitive data, corrupting data, and/or just performing generally illegal or unethical activity

• User must trust the nodes

• Working accurately and effectively• Consideration given to preventing

malicious participants from producing false, misleading, or erroneous results or using the system for ill-intent.

• Authorization levels

• Redundant processing

Design Hazards

• Concern over reliability• Nodes are independent

and dispersed in nature• No way to guarantee the

availability of the resource

• Resource could viably drop out at any time

• Time constraints used to check tasks at an estimated interval. If it does not, the task is immediately reassigned to another node.

When should a grid be considered?

Applications that:• Take a long time to

execute• Process large amounts

of data• Involve many runs of

the same task• Can be broken down

into execution units• Involve data that can

be broken down into data sets

Examples

BOINC

• Berkeley Open Infrastructure for Network Computing

• Middleware• Used in:

• Rosetta@home• SAT@home• SETI@home “BOINC Logo”

Source http://boinc.berkeley.edu/logo/boinc_logo.png

SETI

• Search for Extraterrestrial Intelligence

The “ATA”Source http://msnbcmedia.msn.com/i/msnbc/Components/Photos/071012/071012_telescope_hmed_3p.jpg

Allan Telescope Array

“ATA Map”Source http://archive.seti.org/ata/images/Antenna_map_sm.gif

SETI@home

• Powered by BOINC• 35 GB tapes

• 2.5 MHz range• Broken up into "work-units"

“SETI@home Data”Source http://setiathome.berkeley.edu/

SETI Screensaver Progress

“SETI@home Processing”Source http://setiathome.berkeley.edu/

Novartis

Novartis• Switzerland based global pharmaceutical

company• Grid connects over 3,000 of their employees PCs• Provides independent researchers with larger data

sets and greater computing power• Geared toward drug discovery

Source: http://novartis.com

Entelos

Entelos• Biotechnology firm in California

• Use a Grid for drug discovery and research• Grid is set up by Platform Computing• 125 servers and 20 desktops connected • Simulates thousands of tests on hundreds of

“patients” under different circumstances• Processing takes hours instead of days

Source: http://entelos.com/index.php

World Community Grid

World Community Grid• Funded and Operated by IBM• World’s largest non-profit computing grid benefiting

humanity• Volunteer Driven• First Project- “Human Proteome Folding”• Current Project- “Computing for Sustainable Water”

Source: http://worldcommunitygrid.org

CERN

CERN (European Organization for Nuclear Research)

• Their grid supports The Large Hadron Collider• Underground particle accelerator• Runs experiments on particles• Generates about 27 TB of data each day, all of

which is analyzed for interesting readings• Closed Grid

Future of Grid Computing

• Natural progression from internet• Everyone is already

connected to internet• Networks are improving• Rely on computers for

research and analysis• Technology continues

to improve• Moore’s Law!

• …But how will people interact?

Future of Grid Computing

• Out-of-box with grid-enabled software• Could connect with a data cable• OS registers user onto grid and

records activities• Expand to embrace multiple

media types• Security needs to be

addressed before growth• Otherwise there is

overwhelming potential for consumer market

• Indispensable future resource• Research• Education• Everyday computing

Works Cited

• Online Resources• http://www.gridcafe.org/grid-in-30-sec.html

• http://www.trivadis.com/uploads/tx_cabagdownloadarea/grid_computing_en.pdf

• http://www.gridsummit.com/Articles/Milestones.htm

• http://www.pcmag.com/article2/0,2817,763210,00.asp

• http://www.worldcommunitygrid.org/index.jsp

• http://public.web.cern.ch/public/

• http://net.educause.edu/ir/library/pdf/ELI7010.pdf

• http://www.brighthub.com/environment/green-computing/articles/67601.aspx

• http://www.thepicky.com/tech/examples-of-grid-computing-real-world/

• http://www.2020vp.com/hossein-blog/2012/03/the-role-of-grid-computing-in-21st-century-communications/

• http://www.redbooks.ibm.com/redbooks/pdfs/sg246649.pdf

• http://support.sas.com/rnd/scalability/papers/101948_1204.pdf

• http://www.novartis.com/• http://www.entelos.com/index.php

• http://public.web.cern.ch/public/• http://msnbcmedia.msn.com/i/msnbc/Components/Photos/071012/071012_telescope_hmed_3p.

jpg• http://setiathome.berkeley.edu/

Questions?

Thank you!