The 6th International Conference on Computer Science & Education (ICCSE 2011) August !

"# 2011$ SuperStar %irgo# Singapore

ThC $

Cluster, Grid and Cloud Computing: A Detailed Comparison

Naidila Sadashiv
Dept. of Computer Science and Engineering Acharya Institute of echnology !angalore, India sadashiv"acharya.ac.in

S. # Dilip $umar
Dept. of Computer Science and Engineering %niversity &isvesvaraya College of Engineering '%&CE( !angalore, India dilip)umarsm.uvce"*u*.ernet.in

AbstractCloud computing is rapidly growing as an alterna- tive to conventional computing. However, it is based on models like cluster computing, distributed computing, utility computing and grid computing in general. This paper presents an end-to- end comparison between Cluster Computing, Grid Computing and Cloud Computing, along with the challenges they face. This could help in better understanding these models and to know how they differ from its related concepts, all in one go. t also discusses the ongoing pro!ects and different applications that use these computing models as a platform for e"ecution. #n insight into some of the tools which can be used in the three computing models to design and develop applications is given. This could help in bringing out the innovative ideas in the field and can be e"plored to the needs in the computing world. Index TermsCluster Computing, Grid Computing, Cloud Computing, Computing $odels, Comparison.

applica. tion, service and so on that is re>uired 0ithout huge investment on its purchase, implementation, maintenance rather use it for other significant purpose. +esources are re>uested on.demand

I. I N +,D%C I,N -igh.performance computing '-/C( 0as once restricted to institutions that could afford the significantly e1pensive and dedicated supercomputers of the time. here 0as a need for -/C in small scale and at a lo0er cost 0hich lead to cluster computing. he emergence of cluster platforms 0as driven *y a num*er of academic pro2ects, such as !eo0ulf 345, !er)eley N,6 375, and -/&# 385. he popularity of the Internet and the availa*ility of po0erful computers and high.speed net0or) technologies has changed the 0ay computers are used. Grid computing originated in academia in the mid 499:s 0ith an aim to facilitate users to remotely utili;e idle computing po0er 0ithin other computing centers 0hen the local one is *usy 345. Initially, it only referred to a compute grid and had a rather limited audience. -o0ever, after years of development the grid gained momentum and came to mean an effective 0ay for coordinated resource sharing and pro*lem solving in dynamic, multi.institutional virtual organi;ations. Cloud computing, is a )ind of computing model that came into e1istence around the end of 7::<. It provides a pool of computing resources 0hich the users can access through Internet. he *asic principle of cloud computing is to shift the computing done from the local computer into the net0or) 3=5. his ma)es the enterprise use the resource 0hich includes net0or), server, storage,

0ith out any prior reservation and hence eliminates over. provisioning and improves resource utili;ation. o the *est of our )no0ledge, in the literature, only a fe0 comparisons have *een appeared in the field of computing. In this paper 0e *ring out a complete comparison of the three computing models. +est of the paper is organi;ed as follo0s. he cluster computing, grid computing and cloud computing models are *riefly e1plained in Section II. Issues and challenges related to these computing models are listed in Section III. Section I& compares these three models from different perspectives. Section & discusses pro2ects and appli. cations. ools and simulation environment are part of Section &I. Concluding remar)s is presented in Section &II. II.
-+EE

C ,#/% ING # ,DE?S

In this section, 0e *riefly e1plain the three computing models: cluster, grid and cloud. A. Cluster Computing @or many years supercomputer 0as the leader in the field of computing. !ut due to some of the pro*lems faced in the area of science, engineering, and *usiness, 0hich could not *e
&'(!1!)2))!&'1(!(*11*+26$00 ,2011 IEEE )''

effectively dealt 0ith using supercomputers. hey 0ere replaced 0ith clusters 3A5 0ith the aim of overcoming these pro*lems and also offered a very cheap 0ay for gaining access to potentially huge computing po0er. %efinition& A cluster is a collection of parallel or distri*uted computers 0hich are interconnected among themselves using high.speed net0or)s, such as giga*it Ethernet, SCI, #yrinet and Infini*and. hey 0or) together in the e1ecution of com. pute intensive and data intensive tas)s that 0ould *e not feasi*le to e1ecute on a single computer. Clusters are used mainly for high availa*ility, load.*alancing and for compute purpose. hey are used for high availa*ility purpose as they maintain redundant nodes 0hich are used to provide service 0hen system components fail. he performance of the system is improved here *ecause even if one node fails there is another stand*y node 0hich 0ill carry the tas) and eliminates single points of failure 0ithout any hindrance 3A5. 6hen multiple computers are lin)ed together in a cluster, they share computational 0or)load as a single virtual computer. @rom the users vie0 point they are multiple machines, *ut they function

ThC $

as a single virtual machine. he userBs re>uest are received and distri*uted among all the standalone computers to form a cluster. his results in *alanced computational 0or) among different machines, improving the performance of the cluster systems. ,ften clusters are used mainly for computational purposes, than handling I,.*ased activities. B. Computing Grid

)eeps off the *urden of updating the soft0are to the latest version, licensing and is of course more economical. Some e1ample service providers are Salesforce, Customer +elationships #an. agement'C+#( system and Google Apps.

Grid computing 3C5 com*ines computers from multiple administrative domains to reach a common goal, to solve a single tas), and may then disappear 2ust as >uic)ly. It is analogous to the po0er grid 3<5. ,ne of the main strategies of grid computing is to use middle0are to divide and apportion pieces of a program among several computers. Grid computing involves computation in a distri*uted fashion, 0hich may also involve the aggregation of large.scale cluster computing. *ased systems. he si;e of a grid may vary from small a net0or) of computer 0or)stations 0ithin a corporation to large colla*orations across many companies and net0or)s. Some of the definitions on grid computing are given *elo0: %efinition& !uyya et. al. 3<5 defined grid as a type of parallel and distri*uted system that ena*les the sharing, selection, and aggregation of geographically distri*uted autonomous re. sources dynamically at runtime depending on their availa*ility, capa*ility, performance, cost, and users >uality.of.service re. >uirements. %efinition& Ian @oster 3D5 defined grid as a system that coor. dinates resources 0hich are not su*2ect to centrali;ed control, using standard, open, general.purpose protocols and interfaces to deliver nontrivial >ualities of service. C. Computing Cloud

Cloud computing 395 refers to *oth the applications deliv. ered as services over the Internet and the hard0are and system soft0are in the data centers that provide those services. %efinition& According to !uyya et. al. 34:5 a cloud is a type of parallel and distri*uted system consisting of a collection of interconnected and virtuali;ed computers that are dynamically provisioned and presented as one or more unified computing resources *ased on service.level agreement. Cloud computing is a model for ena*ling u*i>uitous, convenient, on.demand net0or) access to a shared pool of configura*le computing resources 'e.g., net0or)s, servers, storage, applications, and services( that can *e rapidly provisioned and released 0ith minimal management effort or service provider interaction 3445. Cloud computing provides *asically three )inds of ser. vice: Soft0are as a Service 'SaaS(, /latform as a Service 'PaaS( and Infrastructure as a Service 'IaaS.( E SaaS is a )ind of services 0here in many users can ma)e use of the soft0are hosted *y the service provider and pay only for time its *eing used. It 0ill *e *etter than *uying the hard0are and soft0are as
222

PaaS provides a high.level integrated environment to de. sign, *uild, test, deploy and update online custom applica. tions. Some e1ample service providers are GoogleBs App Engine , #icrosoft A;ure , +ightScale and Sales@orce . IaaS refers to the services provided to the users to use processing po0er, storage, net0or) and other computing resources, to run any soft0are including operating sys. tems and applications. Some of the IaaS providers are A6S, Eucalyptus, ,pen Stac), GoGrid and @le1iscale. III. C -A??ENGES IN C ?%S E+ , G +ID AND C ?,%D C ,#/% IN
G

ThC $ interdependent. A( -eterogeneity: @inding 0ays to create a 0ide area data intensive programming and scheduling frame0or) in heterogeneous set of resources. C( /rogramming: he lo0.coupling *et0een nodes and the distri*uted nature of processing ma)e the programming of applications over grids more comple1.

Every computing model has similar or uni>ue challenges, needs to convert challenges into opportunities and has scope for further research. In this section, 0e highlight some of the challenges of the three computing models considered. A. Challenges in the Cluster Computing he research challenges of cluster computing are as follo0s: 4( #iddle0are: o produce soft0are environments that pro. vides an illusion of a single system image, rather than a collection of independent computers. 7( /rogram: he applications that run on the clusters must *e e1plicitly 0ritten 0hich incorporates the division of tas)s *et0een nodes, also the communication *et0een them should *e ta)en care of. 8( Elasticity: he variance in real.time response time 0hen the num*er of service re>uests changes dramatically. =( Scala*ility: o meet the additional re>uirements of a resource thus effecting the performance of the system. B. Challenges in grid computing he research challenges faced in grids include: 4( Dynamicity: +esources in grid are o0ned and managed *y more than one organi;ation 0hich may enter and leave the grid at any time causing *urden on the grid. 7( Administration: o form a unified resource pool, a heavy system administration *urden is raised along 0ith other maintenance 0or) to coordinate local administration policies 0ith glo*al ones. 8( Development: /ro*lems are concerned 0ith 0ays of 0riting soft0are to run on grid.computing platforms, 0hich includes to decompose and distri*ute to process. ing elements, and then assem*ling solutions. =( Accounting: @inding 0ays to support different account. ing infrastructure, economic model and application mod. els that can cope 0ell 0ith tas)s that communicate fre>uently and are
22

C. Challenges in the cloud computing he challenges of cloud computing includes the follo0ing: 4( Dynamic scala*ility: he compute nodes are scaled up and do0n dynamically *y the application according to the response time of the userBs >ueries. he scheduling delays involved are real concern 0hich leads to the need of effective and dynamic load management system. 7( #ulti.tenancy: 6hen the num*er of applications running on the same compute node increases, it 0ill reduce the amount of *and0idth allocated to each application 0hich may lead to performance degradation. 8( Fuerying and access: Scala*le provenance >uerying and secure access of provenance information are open pro*. lems for *oth grid and cloud environment. =( Standardi;ation: As every organi;ation has their o0n A/Is and protocols used 0hich ma)es the user data or vendor loc).in. hus integration and interopera*ility of all the services and application is a challenge. A( +elia*ility and fault.tolerance: ools for testing the ap. plication against fault tolerance and compute failures are re>uired 0hich help in developing a relia*le system. C( De*ugging and profiling: /arallel and remote de*ugging has al0ays *een a pro*lem for developing -/C programs and is an issue in cloud computing also. <( Security and /rivacy: he user has no idea 0here data is stored and 0ho 0ill use it as there are more hac)ers than developers. D( /o0er: hough cloud computing offers many type of services finally to meet the needs of users, enormous amount of po0er is consumed. An autonomic energy a0are resource management is very much re>uired. I&. C ,#/A+IS,N ,@ C?%S E+ , G+ID AND
C?,%D C,#/% ING

Some of the pro2ects in the field of cluster computing are:

4( Condor 3475: It is a multifaceted pro2ect engaged in five primary activities. a( +esearch in distri*uted computing. *( /articipation in the scientific community. c( Engineering of comple1 soft0are. d( #aintenance of production environments. e( Education of students.

&ie0ed in a *road sense, the concepts of cluster, grid and cloud computing seems to have similar features. his section puts light to differentiate in different perspectives and give an end.to.end comparison. It could *e understood easily 0hen represented in a ta*ular form as given in ta*le I. &. / +,GEC S AND A //?ICA I,NS IN C ?%S E+ , G +ID
AND

C ?,%D C ,#/% ING In this section, 0e present some of the pro2ects and appli. cations in cluster, grid and cloud computing. A. Cluster computing projects and applications

A ! ? E I C ,#/A+IS,N ,@ C ?%S E+ , G +ID AND C ?,%D C ,#/% ING Clusters S?A Allocation +esource -andling ?oose coupling /rotocolsIA/I +elia*ility Security %ser friendliness &irtuali;ation Interopera*ility Standardi;ed !usiness #odel as) Si;e S,A #ultitenancy System /erformance Self service Computation service -eterogeneity Scala*le Ine1pensive Data ?ocality E1ploited Application S0itching cost &alue Added Services -/C,- C ?o0 No -/C, - C, !atch S#E interactiv ?o0 -alf -igh Hes ?imited Centrali;ed Centrali;ed No #/I, /arallel &irtual No Hes No -alf Hes Hes No Single large No No Improves No Computing No No No No Grids Hes Decentrali;ed Distri*uted !oth #/I,#/IC-.G, GIS,G+A# -alf -alf half -alf Hes Hes No Single large Hes Hes Improves Hes #a1. Computing Hes -alf No No Clouds Hes !oth !oth Hes C/II/,S, +ES ,AGAJ @ull No Hes Hes -alf No Hes Small K medium Hes Hes Improves Hes ,n demand Hes Hes Hes Hes

7( Hadoop 34=5 is an open.source frame0or) for running data.intensive applications in a processing cluster *uilt from commodity server hard0are. Some customers use -adoop clustering to analy;e customer search patterns for targeted advertising. ,ther applications include fil. tering and inde1ing of 0e* listings, facial recognition algorithms to search for images in a large data*ase.

7(

Sha+CS 'Shared +esearch Computing Services( 3485: his pilot pro2ect has *een designed to define, demon. strate and measure ho0 shared computing and storage clusters residing in regional data centers can provide computing services to investigators.

Some of the applications in the field of cluster computing are: 4( he 6eather +esearch and @orecast '6+@( is a colla*. orative pro2ect *y several institutions to develop a ne1t. generation regional forecast model and data assimilation system for operational numerical 0eather prediction and atmospheric research.

8(

Clusters 0ere also used for solving grand challenging applications such as 0eather modeling, automo*ile crash simulations, life sciences, computational fluid dynamics, nuclear simulations, image processing, electromagnetic, data mining, aerodynamics and astrophysics. Clusters 0ere used as a platform for data mining applications 0hich involve *oth compute and data intensive opera. tions. hey are also used for commercial applications li)e in *an)ing sector to have achieve high availa*ility and for *ac)up. Clusters are used to host many ne0 Internet service sites li)e -otmail, 0e* applications, data*ase, and other commercial applications. Projects and

mammo. gram analysts, specialist radiologists 0or)ing in *reast cancer screening 0ho can use the grid information in. frastructure to resolve common image analysis pro*lems. 7( DDGrid 'Drug Discovery Grid( 37:5: his pro2ect aims to *uild a colla*oration platform for drug discovery using the state.of.the.art /7/ and grid computing technology.

B. Grid applications

Some of the pro2ects in grid computing are *riefed here. 4( Globus 34A534C5: It is an open source grid soft0are that addresses most challenging pro*lems in distri*uted resource sharing. he first release of Glo*us DemoGrid 34<5, a tool that 0ill *uild an instructional grid envi. ronment that can *e deployed using virtual machines on a cloud or on physical resources. he main goal of DemoGrid is to provide easy access to an environment 0ith various grid tools, 0ithout having to install the tools themselves or needing an account on an e1isting grid. 7( EGI-InSPIRE 34D5 pro2ect 'Integrated Sustaina*le /an. European Infrastructure for +esearchers in Europe(: Started in 7:4:, co.funded *y the European Commission for four years, as a colla*orative effort involving more than A: institutions in over =: countries. Its mission is to esta*lish a sustaina*le European Grid Infrastructure 'EGI( to 2oin together the ne0 Distri*uted Computing Infrastructures 'DCIs( such as clouds, supercomputing net0or)s and des)top grids, for the *enefit of user communities 0ithin the European +esearch Area. he ultimate goal of EGI.InS/I+E is to provide European scientists and their international partners 0ith a sustain. a*le, relia*le e.Infrastructure that can support their needs for large.scale data analysis. 8( Some other grid pro2ects are the NS@s National echnol. ogy Grid, NASAs Information /o0er Grid , Gri/hyN, NEESgrid, /article /hysics Data Grid and the European Data Grid. he grid applications range from advanced manufacturing, nu. merical 0ind tunnel, oil reservoir simulation, particle physics research, -igh Energy Nuclear /hysics '-EN/(, 0eather mod. eling, *io.informatics, terrain analysis of nature o*servation, scientific data*ase, and popular science 0e* services. 4( #ammoGrid 3495: It is a service.oriented architecture *ased medical grid application. he aim is to deliver a set of evolutionary prototypes to demonstrate the

his pro2ect intends to solve large.scale computation and data intensive scientific applications in the field of medicine chemistry and molecular *iology 0ith the help of grid middle0are. ,ver 4:C compounds data*ase 0ith 8.D structure and physiochemical properties are also provided to identify potential drug candidates. C. Cloud Applications Projects and

li*raries, it can conserve li*rary resources and improve its user satisfac. tion. Although there are ,/AC ',nline /u*lic Access Catalog( and I?? 'Inter.li*rary loan( services, access to the shared resources through an uniform access platform

Some of the initiative pro2ects in the area of cloud comput. ing are *riefly presented *elo0: . 4( CER : he European organi;ation for Nuclear +esearch is developing a mega computing cloud to distri*ute data to scientists around the 0orld as part of the ?-C pro2ect. 7( !ni"ied Cloud Inter"ace #!CI$: his pro2ect is proposed *y the Cloud Computing Interopera*ility @orum 'CCI@( 0or)ing on the development of standard A/Is 0hich can *e used *y all cloud service providers to over come the interopera*ility issues. 8( Cloud-Enabled Space %eather Plat"orm #CES%P$: he purpose is to *ring the po0er and fle1i*ility of cloud computing to space 0eather physicists. he goal is to lo0er the *arriers for the physicists to conduct their science, i.e., to ma)e it easier to colla*orate 0ith other scientists, develop space 0eather models, run simula. tions, produce visuali;ations and ena*le provenance. =( &pen ebula: his pro2ect is successful 0ith the release of the most advanced and fle1i*le enterprise. ready cloud management tool, and the gro0th of an active open. source community. hey aim to develop the most. advanced, highly.scala*le and adapta*le soft0are tool)it for cloud computing management. hey plan for opera. tion net0or) to simplify the management of ,penNe*ula cloud instances, fault tolerance functionality to ma1imi;e uptime in the cloud, enhanced management of images and templates, ne0 security functionality, enhanced sup. port for federation of data centers and support for multi. tier architectures. A( 'Clouds 3745 : he goal of the pro2ect is to prototype an advanced cloud infrastructure that can deliver a ne0 level of secure, private and resilient computing and storage that is cost. efficient, simple and scala*le. o achieve security, resiliency and scala*ility needed 0hen outsourcing critical I .systems to a cloud, scientists 0ill *uild an advanced Cloud o" Clouds frame0or) for the pro2ect. his frame0or) provides multiple *ac). ups of the Clouds data and applications in case of a hard0are failure or intrusion. Some of the applications that can *e deployed on cloud include 0e* hosting, media hosting, multi.tenant service, -/C, distri*uted storage, multi.enterprise integration, etc. 4( 6ith the e1pansion of cloud computing and esta*lishing a pu*ic cloud among many %niversity

is difficult 0hich can *e made possi*le 0ith the adoption of cloud computing in such li*raries 3775 3785. 7( RoboEarth is a European pro2ect led *y the Eindhoven %niversity of echnology, Netherlands, to develop a 666 for ro*ots, a giant data*ase 0here ro*ots can share information a*out o*2ects, environments, and tas)s. +esearchers at AS,+, 'A. Star Social +o*otics ?a*o. ratory, Singapore( have *uilt a cloud.computing infras. tructure that allo0s ro*ots to generate 8.D maps of their environments much faster than they could 0ith their on*oard computers 37A5. 8( Cloudo, a free computer that lives on the Internet, right in the 0e* *ro0ser. It allo0s to access documents, photos, music, movies and all other files from any computer or mo*ile phone. =( Panda Cloud antivirus 37C5, the first free antivirus from the cloud. +egular updates is not a pro*lem and it occupies very little system resources, uses collective intelligence servers for fast detection, simple interface and protects /C offline. &I. ,,?S AND S I#%?A I,N E N&I+,N#EN In this section, 0e highlight some of the tools and simu. lators that can *e used to develop applications on the three computing models. A. 'ools "or cluster computing 4( imrod 37<5 is a tool for parametric computing on clusters and it provides a simple declarative parametric modeling language for e1pressing a parametric e1peri. ment. !y using this it can *e easy to create a plan for a parametric computing and use the Nimrod runtime system to su*mit, run, and collect the results from multiple cluster nodes. 7( PAR(& 37D5 is a tool that allo0s the monitoring of system resource and their activities at three different levels: system, node and component. It is used to monitor C.DAC /A+A# 4:::: supercomputer, a cluster of =D %ltra.= 0or)stations po0ered *y Solaris. 8( Condor 34753795 is a speciali;ed 2o* and resource man. agement system for compute intensive 2o*s. It provides a 2o* management mechanism, scheduling policy, prior. ity scheme, and resource monitoring and management. %sers su*mit their 2o*s, tool chooses 0hen and 0here to run them *ased upon a policy, monitors their progress, and ultimately informs the user upon completion. =( (PI and &pen(P 38:5: #essage passing li*raries pro. vide a high.level means of passing data *et0een process e1ecuting on distri*uted memory systems. hey help to achieve high performance out of collections of individual cluster nodes. #/I I is the de facto standard for parallel programming on clusters 0hich consists of a rich set of li*rary functions to do *oth point.to.point and collective communication

among parallel tas)s. A( ,ther cluster simulators include )le*i-Cluster a simu. lator for a single compute cluster, +ERI'AS a cluster server simulator, etc.

he tools used in the field of grid computing for resource discovery, management and performance analysis are: 4( Parad,n, a performance analysis tool supports perfor. mance e1periment management through techni>ues for >uantitatively comparing several e1periments and per. formance diagnosis *ased on dynamic instrumentation. E1periments have to *e set up manually, 0hereas per. formance analysis is done automatically. 7( imrod-G 37<5 uses the Glo*us middle0are services for dynamic resource discovery and dispatching 2o*s over computational grids. It allo0s scientists and engineers to model parametric e1periments and transparently stage the data and program at remote sites, run the program on each element of a data set on different machines and finally gather results from remote sites to the user site. 8( Condor-G 3475 represents the 0or) of Glo*us and Con. dor pro2ects 0hich ena*les the utili;ation of large col. lections of resources that span across multiple domains as if they all *elonged to the userBs personal domain. @rom Glo*us comes the use of protocols for secure inter domain communications and standardi;ed access to remote *atch systems. @rom Condor comes the concern of 2o* su*mission, 2o* allocation, error recovery, and creation of a friendly e1ecution environment . =( Globus, an open source soft0are tool)it that facilitates construction of computational grids and grid *ased ap. plications, across corporate, institutional and geographic *oundaries . A( Gridbus 'G+ID computing and !%Siness( 3845 tool)it pro2ect is associated 0ith the design and development of cluster and grid middle0are technologies for service. oriented computing. It provides end.to.end services to aggregate or lease services of distri*uted resources de. pending on their availa*ility, capa*ility, performance, cost, and FoS re>uirements. C( -egion 3875 is an o*2ect.*ased metasystem that supports transparent core scheduling, data management, fault tol. erance, site autonomy, and a middle0are 0ith a 0ide range of security options. <( ,ther simulators used in grid computing includes Grid. sim 3=:5, .E '!RI&, ,ptorsim, Chicsim, etc. B. 'ools used in Cloud Computing &arious tools and products 0hich provide assistance in development applications on cloud computing are: 4( .enoss a single, integrated product that monitors the en. tire I infrastructure, 0herever it is deployed 'physical, virtual, or in cloud(. It manages the net0or)s, servers, virtual devices, storage, and cloud deployments. 7( Spring Roo is a ne1t generation rapid application

de. velopment tool, com*ined 0ith the po0er of Google 6e* ool)it 'G6 ( that ena*les developers to *uild rich *ro0ser apps in enterprise production environments. 8( CloudSim and CloudAnal,st are important for developers to evaluate the re>uirements of large. scale cloud ap. plications in terms of geographic distri*ution of *oth

computing servers and user 0or)loads. he former 0as developed 0ith the purpose of studying the *ehavior of applications under various deployment configurations 0hereas the latter helps developers 0ith insights in ho0 to distri*ute applications among cloud infrastructures and value added services. =( Cloudera 3=45, an open.source -adoop soft0are frame. 0or) is increasingly used in cloud computing deploy. ments due to its fle1i*ility 0ith cluster. *ased, data. intensive >ueries and other tas)s. It allo0s to e1plore comple1, non.relational data in its native form. &II. C ,NC?%SI,N In this paper, 0e have presented a detailed comparison on the three computing models, cluster, grid and cloud computing. he issues and challenges related to these computing models are highlighted. he pro2ects and applications in various fields are *riefly discussed. Also the tools and simulation environ. ments useful for development of applications are highlighted. Such a comparison in different perspectives 0ill ma)e easy to understand the computing models since the features of these computing models seems to *e similar conceptually. It also helps in identifying the similarities and differences from each other. Grid and cloud computing appears to *e a promising model especially focusing on standardi;ing A/Is, security, interopera*ility, ne0 *usiness models, and dynamic pricing systems for comple1 services. -ence there is a scope for further research in these areas. + E@E+ENCE
S 345 !eo0ulf, http/00111.beo1ul".org 375 -igh /erformance &irtual #achine, http/00111csag.ucsd.edu0projects0hp2m.html. 385 . Anderson, D. Culler, and D. /atterson, LA Case for Net0or) of 6or)stationsM, IEEE (icro, 4A'4(, pp. A=.C=, 499A. 3=5 I. @oster, C. $esselman, S. uec)e, L he Grid: !lueprint for a Ne0 Com. puting InfrastructureM, Intl. 3r. o" Supercomputer Applications, 4A'8(, 7::4. 3A5 LCluster ComputingM, http/00en.1i4ipedia.org01i4i0Cluster computing. 3C5 $. $rauter, +. !uyya, and #. #ahes0aran, LA a1onomy and Survey of Grid +esource #anagement Systems for Distri*uted ComputingM, 3r. o" So"t1are Practice and E*perience, 87,'7(, pp. 48A.4C=, 7::7. 3<5 #.Chetty and +. !uyya, L6eaving Computational Grids: -o0 Anal. ogous Are hey 0ith Electrical GridsNM, Computing in Science and Engineering #CiSE$,=, pp. C4.<4, 7::7. 3D5 I. @oster, L6hat is the GridN A hree /oint Chec)listM, 7::7. 395 #. Arm*urst et al., LA*ove the Clouds: A !er)eley &ie0 of Cloud ComputingM, 'ech. report, %niv. of California, !er)eley, 7::9. 34:5 +. !uyya, C. S. Heo, S. &enugopal, G. !ro*erg, and I. !randic, LCloud computing and emerging it platforms: &ision, hype, and reality for de. livering computing as the Ath utility, @uture GenerM, Computer S,stems, 7A'C(, pp. A99.C4C, 7::9. 3445 NIS , http/00csrc.nist.go20groups0S S0cloudcomputing0. 3475 LCondor tool)itM, http/00111.cs.1isc.edu0condor0condorg. 3485 LSha+CSM, http/00srcs.ucop.edu0pilot.php.

34=5 L-adoopM, http/00hadoop.apache.org. 34A5 LGlo*usM, http/00111.globus.org0ogsa0. 34C5 Ian @oster and C. $esselman, LGlo*us: A #etacomputing Infrastructure ool)itM, Intl. 3r. o" Supercomputer Applications, 44'7(, 499<. 34<5 LGlo*us pro2ectM, http:II000.glo*us.orgIdemogridI. 34D5 LEGI.InS/I+EM, http:II000.egi.euIpro2ectsIegi. inspireI. 3495 L#ammoGridM, http/00mammogrid.2itamib.com. 37:5 LDDGridM, http/00111.ddgrid.ac.cn. 3745 LCloud Security Alliance 'CSA(M, http/00111.cloudsecurit,alliance.org0trustedcloud.html.

3775 @an !ingsi, LCloud Computing and ?i*raries: Defense for +esearch on the Cloud ComputingM, 3r. o" -ibrar, and In"ormation Ser2ice, 74, 7::9. 3785 @eng Jiaona, !ao ?ingyun, LApplication of Cloud ComputingM, 5rd Intl. Con". on Ad2anced Computer 'heor, and Engineering #ICAC'E$, 7:4:. 37=5 Ai0u Shi, Houfu Jia and -aiyan Ohan, LApplying Cloud Computing in @inancial Service IndustryM, Intl. Con". on Intelligent Control and In"ormation Processing, 7:4:. 37A5 Erico Gui;;o, L+o*ots 0ith their heads in the cloudM, IEEE Spectrum, 7:44. 37C5 L/anda CloudM, 111.cloudanti2irus.com0 37<5 +. !uyya, D. A*ramson and G. Giddy, LNimrodIG: An Architecture for a +esource #anagement and Scheduling System in a Glo*al Computa. tional GridM, IEEE Intl. Con". on High Per"ormance Computing #HPC ASIA$ 7:::. 37D5 L/A+#,NM, http/00111.cloudbus.org0course0parmon.php. 3795 D. hain, . annen*aum, and #. ?ivny, LDistri*uted computing in, practice: he Condor E1perienceM, Concurrenc, and Computation/ Prac- tice and E*perience, 4<, pp. 878.8AC, 7::A. 38:5 N. $aronis, !. oonen, and I. @oster, L#/IC-.G7: A Grid. Ena*led Implementation of the #essage /assing InterfaceM, 3r. o" Parallel and 6istributed Computing,C8, pp.AA4.AC8, 7::8. 3845 LGrid*us tool)itM, http/00111.gridbus.org. 3875 L?egion tool)itM, http/00111.legion.2irginia.edu. 3885 +. /rodan, . @ahringer, L,n %sing OEN %+I, for /erformance and /arameter Studies on Cluster and Grid ArchitecturesM, Proc. o" the 77th Euromicro Con". on Parallel8 6istributed and et1or4-Based Processing #Euro-P6P9:5$, 7::8. 38=5 @. !erman, A. -ey and G. @o1, LGrid Computing: #a)ing the Glo*al Infrastructure a +ealityM, 3ohn %ile, and Sons, 7::8. 38A5 +adu /rodan and homas @ahringer, LOEN %+I,: An E1periment #anagement System for Cluster and Grid ComputingM, Proc. o" the IEEE Intl. Con". on Cluster Computing #C-!S'ER9:;$. 38C5 ?. #. &a>uero, ?. +odero.#erino, G. Caceres, and #. ?indner, LA *rea) in the clouds: to0ards a cloud definitionM, SIGC&(( Computer Comm. Re2ie1, 89'4(, pp. A:.AA, 7::9. 38<5 ?uis #. &., ?uis +.#, +. !uyya, LDynamically Scaling Applications in the CloudM, AC( SIGC&(( Comp. Comm. Re2ie1, =4'4(, 7:44. 38D5 I. @oster, H. Ohao, I. +aicu and S. ?u, LCloud Computing and Grid Computing 8C:.Degree ComparedM, Proc. IEEE Grid Computing En2i- ronments %or4shop, pp. 4.4:, 7::D. 3895 Hi 6ei and #. !rian !la)e, LService.,riented Computing and Cloud Computing Challenges and ,pportunitiesM, IEEE Comp. Societ,, 7:4:. 3=:5 +. !uyya and #. #urshed, LGridSim: A ool)it for #odeling and Simulation of Distri*uted +esourceM, Concurrenc, and Computation/ Practice and E*periance, 4=, pp. 44<A.477:, 7::7. 3=45 LClouderaM, http/00111.cloudera.com. 3=75 6on $im. LCloud Computing: oday and omorro0M, 3r. o" &bject 'echnolog,, D'4(, 7::9. 3=85 #. G. Chin, S. -arvey, S. Gha, and /. &. Coveney, LScientific Grid Com. puting: he @irst GenerationM, Computing in Science and Engineering, <, pp. 7=.87, 7::A. 3==5 &ecchiola, #. $irley, and +. !uyya, L#ulti.,*2ective pro*lem solving 0ith ,ffspring on Enterprise CloudsM, Proc. 7:th Intl. Con". on High Per"ormance Computing #HPC Asia9:<$, 7::9. 3=A5 +. !uyya 'Ed.(, L-igh /erformance Cluster Computing: Architectures and SystemsM, 4, 4999. 3=C5 !. Allcoc), et. al. LData #anagement and ransfer in -igh /erformance Computational Grid EnvironmentsM, Parallel Computing 3r., 7D'A(, pp. <=9.<<4, 7::7. 3=<5 I. @oster and C. $esselman, LGlo*us: A #etacomputing Infrastructure ool)itM, Intl. 3r. Supercomputer Applications, 44'7(, pp.44A.47D, 499<. 3=D5 I. @oster, C. $esselman and S. uec)e, L he anatomy of the Grid: Ena*ling Scala*le &irtual ,rgani;ationM, Intl. 3r. o" High Per"ormance Computing Applications, 4A'8(, pp. 7::.777, 7::4. 3=95 H. Ohao, I. +aicu, I. @oster, LScientific 6or)flo0 Systems

for 74st Century, Ne0 !ottle or Ne0 6ineNM, IEEE %or4shop on Scienti"ic %or4"lo1s, 7::D. 3A:5 F. Chen and F. N. Deng, LCloud Computing and its $ey echni>uesM, 3r. o" Computer Applications, =, pp. 7A.C7, 7::9.