Comparing Cloud Computing and Grid Computing

Submitted To Prof. Amit Saraswat Rajesh Raghavan Section A Roll No. 43(Old) 20-Nov-12

Objective To Compare Cloud Computing and Grid Computing and deduct which is better for corporate enterprise Methodology A study about 11 different papers has been done in order to compare the two technologies. Literature Review Cloud computing One good definition of Cloud Computing is given by Y.Zhao, M. Wilde, I. Foster, J. Voeckler, J. Dobson, E. Gilbert,T. Jordan and E. Quigg in 2005: A Large-scale distributed computing paradigm that is driven by economies of scale, in which a pool of abstracted, virtualized, dynamically-scalable, managed computing power, storage, platforms, and services are delivered on demand to external customers over the Internet"[1]. Cloud Computing offers its users services and applications, which are provided through the Internet, and thus, a large number of computers will be in the path of the data when the data is sent to the Cloud for analysis, and, also, when the data is coming back from the Cloud, once the analysis has been already carried out. For instance, a Mobile phone or PDA with low processing power and/or low storage capacity could take advantage of the Cloud and store or process data inside it, which would allow the mobile phone or the PDA to use applications which require more processing power than the processing power available in these kinds of electronic devices and, which will also require a permanent broadband connection in order to send and receive data at any moment from any place, thus allowing the user to ac-cess the services wherever you are. As another example of possible users that would be interested in Cloud Computing, there are lots of small and medium-sized companies that want to provide users with a new service which requires a lot of processing power, storage power and even networking capacity. To establish this infrastructure, the company should invest a lot of money in a large number of computers to process all the incoming requests, even if the peak load appears only once in a while and for these companies it makes it impossible to deploy the service due to the lack of resources. To solve this, one very small company which wants to provide a new service to a large number of customers could either use its own private Cloud (which can be either one or two computers, for example) and then, every time a peak demand appears, use the public Cloud to process the data (one example is Amazon EC2) or directly put its server into the Cloud and pay for the amount of processing power or storage capacity used at the end of the month: instead of contracting two servers for the full month, when required, they could contract one small server and increase the processing capacity through the Cloud processing and/or storage services. As stated by David Linthicum [2], Cloud Computing can be given in some different ways: Storage-as-a-service: It offers disk space on demand. Database-as-a-service: Gives the possibility of using remote services which other users might be using at the same time. Information-as-a-service: Allows access to data stored in a remote server. Process-as-a-service: To be able to have processes running in the Cloud.

 Application-as-a-service: It means to have one application provided by the Cloud provider available on demand through the Cloud, for instance, a very heavy application that requires a lot of storage or processing power. Platform-as-a-service: SDK provided by the Cloud provider that allows the programmer to develop new applications that will run in the Cloud. Integration-as-a-service: Allows having an integrated system which acts like an intermediary between the data given by the company and the Software allocated into the Cloud (This software can make some XML parsing or even format conversion). Security-as-a-service: Grants the possibility to have security applications running directly in the Cloud: for instance, an antivirus system. Management-as-a-service: Service that allows the controlling and managing of multiple Clouds from one place. Testing-as-a-service: Allows us to make testing through specific testing software running in the Cloud, for example, testing Artificial intelligence software that requires lots of processing power. Infrastructure-as-a-service: Provides datacentre-as-a-service, which means to have access remotely to computing resources. With cloud computing, companies can scale up to massive capacities in an instant without having to invest in new infrastructure, train new personnel, or license new software. Cloud computing is of particular benefit to small and medium-sized businesses who wish to completely outsource their datacentre infrastructure, or large companies who wish to get peak load capacity without incurring the higher cost of building larger data centres internally. In both instances, service consumers use what they need on the Internet and pay only for what they use. The service consumer no longer has to be at a PC, use an application from the PC, or purchase a specific version that's configured for smartphones, PDAs, and other devices. The consumer does not own the infrastructure, software, or platform in the cloud. He has lower upfront costs, capital expenses, and operating expenses. He does not care about how servers and networks are maintained in the cloud. The consumer can access multiple servers anywhere on the globe without knowing which ones and where they are located.

Grid computing Alternatively we have Grid Computing, which is well de-fined by Mary Humphrey and Mary R. Thomson: A Computational Grid is a collection of heterogeneous computers and resources spread across multiple administrative domains with the intent of providing users easy access to these resources.[3].Ian Foster [4] made a three point checklist to define what a Grid is. In this list he said: 1. Coordinates resources that are not subject to centralized control: Which means that not only one entity manages all the system but some different sys-tem administrators could be managing different parts of the same Grid at the same time. 2. Using standards, open, general-purpose protocols and interfaces: This will allow to all the companies involved in the Grid to use and access these standards. 3. To deliver nontrivial qualities of service: In Grid computing not a fixed rate of load is going to be man-aged and this can be given small or big rates. This phenomenon causes the given quality of service to vary and not always stay constant.

Grids allow the use of idle resources. Through this, companies create a Grid in order to share those idle resources and, if necessary, they can access more computational resources (shared by other

companies) than they usually can, and share their own resources while they are not carrying out any computationally demanding tasks. About the architecture, a lot of heterogeneous hardware is used in order to create the Grid and, in addition, these de-vices are not managed by only one person but by different system administrators in each of the companies. This situation causes the security, administration policies and network managing to become heterogeneous too, thus more difficult to manage. However, not only companies can make use of these Grids; users can do it too. One successful example would be the Seti@home project, a program that attempts to find extra-terrestrial life, where users install a program and, while not using the resources, idle resources are used to process small packets or tasks, and return the processed information to the main server. Cloud computing evolves from grid computing and provides on-demand resource provisioning. Grid computing may or may not be in the cloud depending on what type of users are using it. If the users are systems administrators and integrators, they care how things are maintained in the cloud. They upgrade, install, and virtualize servers and applications. If the users are consumers, they do not care how things are run in the system. Grid computing requires the use of software that can divide and farm out pieces of a program as one large system image to several thousand computers. One concern about grid is that if one piece of the software on a node fails, other pieces of the software on other nodes may fail. This is alleviated if that component has a failover component on another node, but problems can still arise if components rely on other pieces of software to accomplish one or more grid computing tasks. Large system images and associated hardware to operate and maintain them can contribute to large capital and operating expenses. Feature Comparison

There are many important features that are common to both Cloud and Grid. Table 2 below presents various features of Grids and Clouds, highlighting the similarities and differences between both paradigms. The first one of these features is Resource Sharing; Grids appear to be fairly sharing resources across organizations, whereas Clouds provide the resources that the Service Provider requires on demand. Another feature is Heterogeneity, both Cloud and Grid Computing support the aggregation of heterogeneous resources. Add to these, Virtualization feature that covers both, data and computing resources [5]. Mutually, Clouds and Grids add the virtualization of hardware resources.
As far as the Security feature is concerned, it has been argued by some authors that security has not been seriously explored [6]. Grids have not dealt with end user security; however in Clouds each user has an access to its individual virtualized environment. Clouds system is facing a serious problem caused by lack of high level services; this may be as a result of the low level of maturity associated to Clouds paradigm. In contrast, Grids have a number of these high level services for instance data transfer, metadata search [7].

Moreover, still there are some valuable features (e.g. Scalability and Self-Management, Usability, Standardization and Payment Model Quality of Service) which are summarized in the table below. Feature Resource Sharing Resource Heterogeneity Virtualization Security High Level Services Architecture Software dependencies Platform Awareness Software Workflow Scalability Self-management Centralization Degree Usability Standardization User access Payment model QoS Guarantees Grid Collaboration(Vos, fair share) Aggregation of heterogeneous resources Virtualization of data and computing resources. Security through credential delegations. Plenty of high level services. Service Oriented. Application domain dependent software. The client software must be Grid enabled. Applications require a predefined workflow of services Nodes and sites scalability Reconfigurability Decentralized Control. Hard to manage Standardization and interoperability. Access transparency for the end user. Rigid Limited support, often besteffort only. Cloud Assigned resources are not shared Aggregation of heterogeneous resources. Virtualization of hardware and software platforms. Security through isolation. No high level services defined yet. User chosen architecture Application domain independent software. The SP software works on a customized environment Workflow is not essential for most applications. Nodes, Sites and Hardware scalability. Nodes, Sites and Hardware scalability. Centralized control (until now) User friendliness Lack of standards for Clouds interoperability. Access transparency for the end user. Flexible Limited support, focused on availability and uptime.

There are many similarities between Grid and Cloud computing, both systems share the same basic goal [8]: to reduce the cost of computing, increase reliability, and increase flexibility by transforming computers from something that we buy and operate our-selves to something that is operated by a third party". However, there are some differences Grids come out from existing heterogeneous resources to add new abstraction layers, allowing the definition of new and complex services by aggregation. In contrast, Clouds are not effectively composed of homogenous components, and it is possible to be designed on top of an existing grid. The essential objective for creating the Cloud system is to provide a determined set of capabilities to the user; therefore the principle of the design is a specific goal interface. In contrast, Grids are considered to be general purpose, so they display a complete set of available system capabilities and the resulting interface available for users and applications remains low level [9].

As stated earlier, Cloud and Grid Computing are similar technologies that share lots of issues. The most important things that both technologies have in common 1. To achieve good scalability, data must be distributed over many computers [10]. 2. People can be afraid of sending sensitive data through a large number of computers.[10]. 3. Data must be moved repeatedly to distant computers, which generates the bottleneck of the process, since the data is not always available everywhere and sometimes it is necessary to make this data available [10] 4. Data can be requested regardless of its location [10]. 5. Cloud and Grid computing provide service-level agreements (SLAs) for guaranteed uptime availability of, say, 99 percent. If the service slides below the level of the guaranteed uptime service, the consumer will get service credit for receiving data late. [11] 6. Both systems must be able to determine the amount of unused resources [11]. Differences between Cloud and Grid Computing Here we have a list of differences between Grid and Cloud in order to clarify what are the main differences between Cloud and Grid computing: 1. Cloud computing normally runs in a set of homogeneous computers, but Grid, on the other hand, runs on heterogeneous computers [10]. 2. Grid computing is normally focused on an intensive calculus, while Cloud Computing offers two types of calculuss: standard and intensive. [11]. 3. Grid computing is open-source while Cloud Computing is not.[7]. This creates interoperability problems between todays Clouds [10]. 4. Most Grids use a batch-scheduled compute model, while in Cloud Computing all users share all the re-sources at the same time. And this is why latency sensitive applications, which could run in Grids, could not be supported in Cloud Computing [10]. 5. Grids do not rely on virtualization as much as Cloud does [10]. 6. Cloud Computing does not support as much provenance as Grid does [10], which is a derivation history of a data product, including all the data sources, intermediate data products, and the procedures that were applied to produce the data product [10]. 7. High Performance computing is better supported in Grid computing than in Cloud Computing [10]. 8. Grid is distributed, has multiple research user communities (including users accessing resources from varied administration domains) and is grouped in Virtual Organizations [7]. Cloud computing, on the other hand, usually has only one research community and a common group of system administrators that take care of the entire Cloud. 9. Grid is mostly publicly funded at local, national and international levels [7], while Cloud Computing is funded mainly by its users.

10. Grid computing, must share user and resource inter-faces to allow providers to connect their resources, while Cloud Computing attempts to share only the user interface while the resources interfaces are hidden and given in an abstract way. 11. In Grid computing, the trust model is based on identity delegation, where users can access and browse resources, which are not highly abstracted and virtualized at different Grid sites. In Cloud computing resources are abstracted and virtualized and this trust model does not exist since everything is inside the same Cloud. 12. While in Grid computing, storing very small files (e.g.1-byte files) is not economically suitable, in Cloud Computing it is. [11]

THE COST OF A CLOUD In this section the discussion focuses on cost models in some Components in Cloud architecture. The table below provides a rough guide to associated costs [12].

Amortized Cost ~45% ~25% ~15% ~15% Server Cost

Component Servers Infrastructure Power draw Network Table 2

Sub Components CPU, memory, storage systems Power distribution and cooling Electrical utility costs Links transit, equipment

As shown in table2 above, the greatest cost goes to servers. There are some structural factors that prevent accomplishing high utilization of data centre. 1. Uneven Application Fit, incorporating unnecessary components such as memory, network and storage, which might be unutilized by applications, fitted on the server.

2. Uncertainty in Demand Forecasts, service demands in Cloud particularly new service need to be high state of certainty by going beyond what forecasts would predict, otherwise it might spike quickly. 3. Long provisioning time scales, purchases have a tendency to be large, with components bought in bulk. Infrastructure is intended to extend over long time periods. Servers are intended to continue as many years as possible, with annual growth ordered [12]. In addition, there are some other valuable features (e.g. Risk Management, Hoarding and Virtualization short-falls), which reflect badly utilization of data centre [12]. Thus, the main effective factor to raise utilization and solve the above problems is agility, the ability to increase and decrease the number of resources to match demand and to extract those resources from optimal location [12].

Infrastructure Cost. As table 1 represents the aggregate cost is substantial. In addition the cost of the difference in time between ordering and delivering is 8 months. However, relaxing the requirement for individual server can be used to reduce the infrastructure cost [12]. Power cost. The metric provided by the green Grid [13] is used to describe the data centre Power usage efficiency (PUE) as PUE= (Total Facility Power). A PUE of the art is 1.7, which is below the average of the normal state. To estimate power draw for a mega data centre, a PUE of 1.7 is assumed, a reasonable utility price of $.07 per KWH, 50,000 servers with each drawing on average 180W (servers draw as much as 65% of peak when idle), the total cost comes to 50, 000 180/1000 1.7 $0.07 24 365 = $9.3 million a year. However, the largest impact on the power cost of data centres is decreasing the power draw of each server. Hardware innovation, including use of high efficiency power supplies and voltage regulation modules are an important improvement to reduce the power cost. In addition, another approach, which has been proposed by Barroso and Hlzle [13], is based on energy proportionality to create servers that are closer to implementing energy proportionality to improve efficiency. Network cost. The networking costs are based in wide area networking [12]: 1. Peering, in delivering packets to end-users. 2. The inter-data centre links carrying traffic between geographically distributed data centres. 3. Regional facilities (backhaul, metro-area connectivity, co-location space) needed to reach wide area network interconnection sites.

Reducing the cost of wide area networking can be achieved using a clever design of peering and transit strategies, combined with optimal placement of micro and mega data centres. Moreover, a better design of the services themselves partitioning their functionality and their state between data centres can be also used to reduce the cost. Economic characteristic Business model The business model is the key characteristic to distinguish grid computing and cloud computing. The grid computing is mainly supported by government and academe. On the one hand, this determines the nature of grid computing: the impulse of profit is not strong enough. On the other hand, the grid computing is a research for future development of information technology. But the cloud computing is mainly supported by gigantic IT companies. They plan that all investments on cloud computing should get return on investment (ROI) in the near future or beat market competitors in the long run. There are many business models especially how-to-pay models in cloud computing. Pay-per-use may be the favourite one in many cases. This is almost the same as the concept of utility computing. The capacity of processing, storage and network in cloud computing is utility service as water, electricity and gas in society [15]. These utility services can be available whenever the user requires them at any time in modern human society. Users pay service providers based on their usage of these utility services. There are two categories of cloud users: end user and median user. Cloud services are ends in themselves for end user. End user consumes cloud services for self-use. Median user consumes cloud

services and cost efficiently supplies professional services to others. End user sometimes doesnt pay for cloud services directly. For example, online game players pay for special game according to how long they stay online. And part charge is defrayed to maintain the running of cloud system. This process is opacity to end user. Median user usually pays for consumed cloud services directly. They save money on jumping to the market quickly. For median user, it is no need to manage complex hardware and software, learn how to use tools and gain experience with cloud computing technology. But the business of cloud computing is far beyond these. There are many free accesses to cloud computing. It is free to search with Google, send email with Hotmail and find new friends with Facebook. These conditions rarely appear in using other social utility services. Maybe air is an exception but clean air may be not free anymore in the future. We still dont know how long we can breathe freely due to air pollution. The strategy of sustaining these free cloud services is out of the scope of this paper. Attention economy may be the key idea. User experience characteristic Ease of Use User experience which belongs to the subject of human computer interaction is an important criterion when evaluating whether an application is successful or not. In cloud computing, user experience improves a lot than its ancestors like grid computing. The cloud service is a means toward the end of providing a good experience for cloud user. The valuable services should be easily accessed by cloud user. The core of user experience is achieving ease use. Ease use is not only simple but also elegant. There are three reasons why cloud computing should be ease use: 1. Most cloud providers offer Internet-based interfaces which are simpler than other application program interfaces (API). These interfaces are simple and elegant enough to hide the business processing behind. The interfaces can stay the same ignoring whether the business processing has changed or not. 2. User experience of web applications is full studied. So the user interfaces are independent of content. The development of web application has a full suite of flow which can be divided into three stages including user need analysis, function design and program implementation. In top-down method, the user experience design is the fundamental of whole function design. The facets of the user experience are useful, usable valuable, desirable, findable, credible and accessible [16]. 3. The web 2.0 increases the interactions between web users and providers. The web was originally designed to transport hypertext. As the rapid and rich developments of increasingly sophisticated contents are appearing, web is usually used as a remote software interface. The web 2.0 is supposed to be the continuum of user experience and blurs the line between software and the Internet. The emerging AJAX technology makes web applications and services are becoming more software-like [17]. All these reduce information technology overhead for the end user. Search in the Internet is very simple: a web page with an input text and a confirm button.

CONCLUSION In conclusion, cloud computing as mentioned above is a new technology of computer network, providing the web services in high quality and lower cost comparing to normal technique. Using cloud computing might contribute to improvement of services in other related technologies specially the previous generations such as Grid computing. Cloud computing is almost certainly set to be developed and become an attractive option for many corporations. The future of computing is set to be successful. It is the next generation and can provide tremendous value to companies. It can help companies achieve more efficient use of their IT hardware and software investments, which in turn can lead the acceleration and adoption of innovations. Clouds consist of data centers which are owned by the same institute. The homogeneity within each data center in the infrastructure is the main feature for the cloud computing compared to grid computing. In this case, any conflict between a heterogeneous data center and/or different administration domains can become a serious issue for cloud interoperability. It can be seen that the stages of anonymity and privacy provided by cloud to the external users will be less than the user of desktop in numerous situations. On the other side, grids were originally established with the knowledge that resources infrastructure are dynamic and heterogeneous in their nature. This implies different organization with different administrative domains. This also means that security was taken into account from the beginning when the grid system was originally built. Presently, the security paradigm for clouds appears to be less secure than the model in grids environment. It has been arguably said that the cloud computing might reduce the cost of the infrastructure by shrinking the number of resources such as servers. Nevertheless, the services that provided in such environments should be working as if in Grid computing for example. Cloud computing enables innovation by minimizing need for innovators to find resources to develop, test, and make their innovations available to the users community. Innovators become free to focus on innovating rather than the logistics and management of systems. The reduced cost of infrastructure will make it a way forward for many organizations. Cloud Architectures shows how flexible is building applications on on-demand infrastructures. GrepTheWeb is a good example, demonstrating this flexibility. The Cloud seems to be the brightness future of many IT infrastructures and systems.

References [1] Trusted Computing Group. TCG https://www.trustedcomputinggroup.org. [2]David Linthicum. Cloud Computing,Deep Dive. http://www.scribd.com/doc/26495704/Cloud-Computing-Deep-Dive-Report. 2009. MPI-SWS

[3] Judith M. Myerson . Cloud computing ver-sus grid computing. March 2009. EECS Department. University of California, Berke-ley.Technical Report No. UCB/EECS-2009-28. http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-28.html. [4] Ian Foster. What is the Grid? A Three Point Check-list. July 2002. Argonne National Laboratory & University of Chicago http://dlib.cs.odu.edu/WhatIsTheGrid.pdf. [5] Members of EGEE-II. An egee comparative study: Grids and clouds - evolution or revolution. Technical report, Enabling Grids for E-sciencE Project, June 2008. Electronic version available at https://edms.cern.ch/document/925013/ [6] S. Jha, A. Merzky, G. Fox. (2008 ). Using clouds to provide grids higher-levels of abstraction and explicit support for usage modes.Technical report, Open Grid Forum. Available at http://grids.ucs.indiana.edu/ptliupages/publications/cloud-grid-saga.pdf. [7] H. Stockinger. (2007). Defining the grid: a snapshot on the current view. The Journal of Supercomputing, (1):317 [8] I. Foster's. (2008). weblog. http://ianfoster.typepad.com/blog/2008/01/theres-grid-in.html [9] S . Jha, A. Merzky, G. Fox. Using Clouds to Provide Grids Higher-Levels of Abstraction and Explicit Support for Usage Modes. [Online] Available: http://www.ogf.org/OGF_Special_Issue/cloudgrid-saga.pdf [10] R. Desisto, D. Plummer, Smith (2007). Tutorial for Understanding the Relationship Between Cloud Computing and SaaS. http://www.gartner.com/DisplayDocument?ref=g_search\&id=640707 [11] G. Boss, P. Malladi, D. Quan, L. Legregni,H. Hall. (2007). Cloud Computing www.ibm.com/developerworks/websphere/zones/hipods/ [12] A. Greenberg, J. Hamilton, D. A. Maltz, P. Patel. (2009). http://ccr.sigcomm.org/online/files/p68v39n1o-greenberg.pdf [13] The Green Grid. URL http://www.thegreengrid.org. [14]L. A. Barroso and U. Hlzle. (2007). "The case for energy-proportional computing," Computer, vol.40, no. 12, pp. 33-37 [15] J. Varia. (2008).Cloud Architectures. http:// jineshvaria.s3.amazonaws.com/public/cloudarchitectures -varia.pdf [16] Werner, J., Geronimo, G., Westphall, C., Koch, F., Freitas, R.: Simulator improvements to validate the green cloud computing approach. In: Network Operations and Management Symposium(LANOMS), 2011 7th Latin American. pp. 1 8 (Oct 2011).

[17] Douglas O. Balen, C.B.W., Westphall, C.M.: Experimental Assessment Of Routing For Grid And Cloud. The Tenth International Conference On Networks (Icn 2011) Pp. 341346 (2011). [18] S. Jha, A. Merzky, G. Fox. (2008 ). Using clouds to provide grids higher-levels of abstraction and explicit support for usage modes.Technical report, Open Grid Forum. Available at http://grids.ucs.indiana.edu/ptliupages/publications/cloud-grid-saga.pdf. [19] The Green Grid. URL http://www.thegreengrid.org. [20] De Chaves, S.A., Uriarte, R.B., Westphall, C.B.: Toward an architecture for monitoring private clouds. IEEE Communications Magazine 49(12), 130137 (2011), http://dblp.uni-trier.de/db/journals/cm/ cm49.html#ChavesUW11. [21] Mell, P., Grance, T.: The nist definition of cloud computing ( draft ) recommendations of the national institute of standards and technology. Nist Special Publication 145(6), 7 (2011), http://csrc.nist.gov/publications/ drafts/800-145/Draft-SP-800-145_cloud-definition.pdf. [22] Sotomayor, B., Montero, R.S., Llorente, I.M., Foster, I.: Virtual infrastructure management in private and hybrid clouds. IEEE Internet Computing 13, 1422 (September 2009), http://dl.acm.org/citation.cfm?id= 1638588.1638692 [23] Werner, J., Geronimo, G., Westphall, C., Koch, F., Freitas, R.: Simulator improvements to validate the green cloud computing approach. In: Network Operations and Management Symposium (LANOMS), 2011 7th Latin American. pp. 1 8 (oct 2011). [24] Douglas O. Balen, C.B.W., Westphall, C.M.: Experimental Assessment Of Routing For Grid And Cloud. The Tenth International Conference On Networks(Icn 2011) Pp. 341346 (2011). [25] Delaney, K.J., Vara, V.: Google Plans Service To Store Users Data (November 27, 2007), Http://Online.Wsj.Com/Article/ Sb119612660573504716.Html [26] Jaeger, P., Lin, J., Grimes, J.: Cloud Computing And Information Policy: Computing In A Policy Cloud? Journal of Information Technology Politics 5(3), 269 283 (2008), Http://Www.Informaworld.Com/Openurl?Genre=Article&Doi=10.1080/19331680802425479&Magi c=Crossref [27] Foster, I., Kesselman, K.: The Grid: Blueprint For A Future Computing Infrastructure. In: Morgan Kaufmann in Computer Architecture And Design (1999). [28] Grimshaw, A.S., Humphrey, M.A., Natrajan, A.: A Philosophical And Technical Comparison Of Legion And Globus. Ibm J. Res. Dev. 48, 233254 (March 2004), Http://Dx.Doi.Org/10.1147/Rd.482.0233 [29] P. Groth, S. Miles, W. Fang, S. Wong, K. Zauner, L. Moreau. Recording and using provenance in a protein compressibility experiment, in Proceedings of the 14th IEEE Int. Symposium von High Performance Distributed Computing (HPDC), 2005. [30] Hadoop, hadoop.apache.org/core/, 2008.

[31] M. Isard, M. Budiu, Y. Yu, A. Birrell, D. Fetterly. Dryad: Distributed Data-Parallel Programs from Sequential Building Blocks, Euro. Conf. on Computer Systems (EuroSys), 2007. [32] N. Karonis, B. Toonen, and I. Foster. MPICH-G2: A Grid-Enabled Implementation of the Message Passing Interface. Journal of Parallel and Distributed Computing, 2003. [33] G.v. Laszewski, M. Hategan, D. Kodeboyina. Java CoG Kit Workflow, Taylor, I.J., Workflows for eScience, 2007. [34] C. Lin, S. Lu, Z. Lai, A. Chebotko, X. Fei, J. Hua, and F. Fotouhi, Service-Oriented Architecture for VIEW: a Visual Scientific Workflow Management System, IEEE International Conference on Services Computing (SCC), pp.335-342, 2008. [35] B. Ludscher, I. Altintas, C. Berkley, D. Higgins, E. Jaeger, M. Jones, E. Lee, J. Tao, Y. Zhao, Scientific workflow management and the Kepler system. Concurrency and Computation: Practice and Experience, 18(10). [36] M.P.I. Forum. MPI: A Message-Passing Interface Standard, 1994. [37] H. Nakada, Y. Tanaka, S. Matsuoka, S. Sekiguchi. Ninf-G: a GridRPC system on the Globus toolkit, Grid Computing: Making the Global Infrastructure a Reality, pp. 625--638, 2003. [38] Open provenance model http://twiki.ipaw.info/bin/view/Challenge/OpenProvenanceModelReview, 2008. [39] The provenance challenge series: http://twiki.ipaw.info/bin/view/Challenge/, 2008. [40] I. Raicu, Y. Zhao, C. Dumitrescu, I. Foster, M. Wilde. Falkon: a Fast and Light-weight tasK executiON framework, IEEE/ACM SuperComputing 2007. [41] I. Raicu, Y. Zhao, I. Foster, A. Szalay. Accelerating Large-scale Data Exploration through Data Diffusion, International Workshop on Data-Aware Distributed Computing 2008. [42] I. Raicu, Z. Zhang, M. Wilde, I. Foster, P. Beckman, K. Iskra, B. Clifford. Towards LooselyCoupled Programming on Petascale Systems, IEEE/ACM Supercomputing 2008. [43] J. M. Schopf, I. Raicu, L. Pearlman, N. Miller, C. Kesselman, I. Foster, M. DArcy. Monitoring and Discovery in a Web Services Framework: Functionality and Performance of Globus Toolkit MDS4, Technical Report, Argonne National Laboratory, MCS Preprint #ANL/MCS-P1315-0106, 2006. [44] J. Silvestre. Economies and Diseconomies of Scale, The New Palgrave: A Dictionary of Economics, v. 2, pp. 8084, 1987. [45] Y. Simmhan, Beth Plale, Dennis Gannon. A survey of data provenance in e-science, SIGMOD Record 34(3): 31-36, 2005. [46] A. Szalay, A. Bunn, J. Gray, I. Foster, I. Raicu. The Importance of Data Locality in Distributed Computing Applications, NSF Workflow Workshop 2006. initiative,

[47] W. C. Tan. Provenance in databases: Past, current, and future, IEEE Data Eng. Bull., 30(4):3 12, 2007. [48] D. Thain, T. Tannenbaum, M. Livny, Distributed Computing in Practice: The Condor Experience Concurrency and Computation: Practice and Experience, Vol. 17, No. 2-4, pages 323356, February-April, 2005. [49] Twenty Experts Define Cloud Computing, SYS-CON Media Inc, http://cloudcomputing.syscon.com/read/612375_p.htm, 2008. [50] What is Cloud Computing? Whatis.com. [51] N. Wilkins-Diehr: Special Issue: Science Gateways - Common Community Interfaces to Grid Resources. Concurrency and Computation: Practice and Experience 19(6): 743-749 (2007) [52] Y. Zhao, M. Wilde, I. Foster, J. Voeckler, J. Dobson, E. Gilbert, T. Jordan, E. Quigg. Virtual Data Grid Middleware Services for Data-Intensive Science, Concurrency and Computation: Practice and Experience, 2005.