A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-area Nile: a case study of a wide-area data-parallel application. In Proc. 18th International Conference on Distributed Computing Systems, pages 506--515, Amsterdam, The Netherlands, May 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....than agreement for most applications. However, the services comprising the Grid software infrastructure itself are long running and are key for sustaining VO activities. Although inexpensive approaches like eventual update combined with randomized scheduling have proven sufficient so far (e.g. see [3]) we believe that a number of Grid services are likely candidates for agreement protocols. The Grid is currently being build as a concerted effort among many institutions and is already supporting leading scientific applications. In this paper we have identified several differences between the ....

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. In Proceedings of the 18th International Conference on Distributed Computing Systems (ICDCS), Amsterdam, Netherlands, pages 506--515, May 1998.

.... 98] Su 99] Weissman 95] Weissman 98] Zhu 98] This observation is the main motivation behind monitoring systems [Wolski 98] Wolski 99a] Ranganathan 96] Lowekamp 98] Although sometimes it is possible to formulate a scheduling problem in a way that can be solved in polynomial time (e.g. Amoroso 98] most instances of scheduling are NP Hard problems. Therefore, most application schedulers use heuristics to navigate through the space of possible schedules. The main components of such heuristics are (i) how two schedules are compared, and (ii) how the space of schedules is traversed. Most ....

....schedulers use heuristics to navigate through the space of possible schedules. The main components of such heuristics are (i) how two schedules are compared, and (ii) how the space of schedules is traversed. Most application schedulers use a performance model to compare two possible schedules [Amoroso 98] Andersen 98] Berman 96] Casanova 96] Ranganathan 96] Shao 97] Su 99] Weissman 95] Others have devised a mechanism to rank schedules without actually estimating the application s performance [Lowekamp 98] Zhu 98] Since application schedulers that rely on performance models provide an ....

Alessandro Amoroso, Keith Marzullo, and Aleta Ricciardi. WideArea Nile: A Case Study of a Wide-Area Data-Parallel Application. ICDCS'98 -- International Conference on Distributed Computing Systems. May 1998.

.... [100] 104] This observation has motivated the appearance of systems that probe grid resources and forecast their availability, as with NWS [101] 102] Komodo [77] and Remos [68] Although sometimes it is possible to formulate a scheduling problem in a way that can be solved in polynomial time [3], most instances of scheduling are NP Hard. Therefore, most application schedulers use heuristics to navigate through the space of possible schedules. The main components of such heuristics are (i) how two schedules are compared, and (ii) how the space of schedules is traversed. Infrastructure ....

.... Discovery Resource Selection Schedule Planning and Performance Model ing Decision Model accessible resources feasible resource sets evaluated schedules best = self scheduling application 108 Most application schedulers use a performance model to compare two possible schedules [3] [4] 9] 19] 77] 80] 89] 99] Others have devised a mechanism to rank schedules without actually estimating the application s performance [68] 104] Since application schedulers that rely on performance models provide an estimate on the application s execution time, they can be more easily ....

Alessandro Amoroso, Keith Marzullo, and Aleta Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. ICDCS'98 -- International Conference on Distributed Computing Systems. May 1998.

....14] are becoming an increasingly important platform for applications which perform calculations over large datasets. Such applications include image acquisition and processing calculations, digital library searches, high performance massive data assimilation, distributed data mining and others [11, 17, 15, 10, 1, 24, 3]. Aggregating distributed resources presents the opportunity to employ or acquire data from very large datasets which are too large to be stored at a single site. This research was supported in part by NASA Graduate Student Research Grant #NGT 2 52251, Department of Defense Modernization ....

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. In ICDCS'98 - International Conference on Distributed Computing Systems, 1998.

....[18] are becoming an increasingly important platform for applications which perform computations over large datasets. Such applications include image acquisition and processing computations, digital library searches, high performance massive data assimilation, distributed data mining and others [3, 6, 16, 17, 20, 25, 35]. Aggregating distributed resources on the grid presents the opportunity to employ or acquire data from massive datasets which are too large to be stored at a single site. This research was supported in part by NASA Graduate Student Research Grant #NGT 2 52251, Department of Defense ....

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. In ICDCS'98 - International Conference on Distributed Computing Systems, 1998.

....and researchers also have access to supercomputer time. The problem of scheduling large distributed applications continues to be investigated for applications which execute in batch scheduled metacomputing environments (e.g. BDG 98, MFS94, WK94] as well as interactive environments (e.g. AMR98, BWF 96, SBWS99, AMR98, AYIE98, Wei98] The work reported herein extends the target domain for PTomo by targeting both parallel supercomputers and interactive resources simultaneously. We have learned three interesting lessons about metacomputing in general as a result of this effort. First, ....

....access to supercomputer time. The problem of scheduling large distributed applications continues to be investigated for applications which execute in batch scheduled metacomputing environments (e.g. BDG 98, MFS94, WK94] as well as interactive environments (e.g. AMR98, BWF 96, SBWS99, AMR98, AYIE98, Wei98] The work reported herein extends the target domain for PTomo by targeting both parallel supercomputers and interactive resources simultaneously. We have learned three interesting lessons about metacomputing in general as a result of this effort. First, the interface exported by ....

Alessandro Amoroso, Keith Marzullo, and Aleta Ricciardi. Widearea nile: A case study of a wide-area data-parallel application. ICDCS'98 International Conference on Distributed Computing Systems, May 1998.

....performance. 1 1.0 Introduction Increasingly, applications require access to remote data sources including digital libraries, scientific databases, images, and movies, located on the Internet or across a wireless network. These applications range from Web browsers to data intensive applications [1][3] to distributed supercomputing applications. Newly emerging metacomputing or grid environments promise to support many of these applications. These systems allow applications to access both geographically separated computing and data resources with the goal of high performance [5] 7] 17] We ....

A. Amoroso, "Wide-area Nile: A Case Study of a Wide-area Data-parallel Application," The 18th International Conference on Distributed Computing Systems, 1998.

....(per slave) 80,60,40,25,20,15 K tasks (master chunks) static 314.84 375.91 452.47 517.07 dynamic 277.82 338.06 357.98 365.79 10 4. 0 Related Work A number of related projects are exploring the scheduling of applications consisting of independent tasks in distributed systems including [2][4] 5] 6] 7] The Nile project [2] is building a high performance wide area computing environment for high energy physics applications. The objective of Nile is to allocate tasks (or jobs) to computers near a dependent data source. Nile is concerned more with job throughput as opposed to ....

....K tasks (master chunks) static 314.84 375.91 452.47 517.07 dynamic 277.82 338.06 357.98 365.79 10 4. 0 Related Work A number of related projects are exploring the scheduling of applications consisting of independent tasks in distributed systems including [2] 4] 5] 6] 7] The Nile project [2] is building a high performance wide area computing environment for high energy physics applications. The objective of Nile is to allocate tasks (or jobs) to computers near a dependent data source. Nile is concerned more with job throughput as opposed to application finishing time. EveryWare is a ....

A. Amoroso, "Wide-area Nile: A Case Study of a Wide-area Data-parallel Application," The 18th International Conference on Distributed Computing Systems, 1998.

No context found.

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-area Nile: a case study of a wide-area data-parallel application. In Proc. 18th International Conference on Distributed Computing Systems, pages 506--515, Amsterdam, The Netherlands, May 1998.

....ignored the differences between Sarah and Bob s requirements or has gravitated towards Bob s since they are more technically challenging. Moreover, by committing to a particular Grid Computing infrastructure (e.g. Condor pools [11] Globus resources [7] Computational Co ops [4] or Nile farms [1]) both Sarah and Bob must restrict their application to utilize only processors available through such an infrastructure. Bob may be happy to do this, but Sarah might have access to many other processors elsewhere. And, since her application is coarse grained, using whatever processors she can ....

.... comprehensive support for grid computing, a goal that is starting to be pursued commercially by companies such as Entropia [6] Other Grid Computing projects have more focused goals, targeting high throughput applications (as Condor [11] and high energy physics applications (as Nile [1]) for example. Yet other efforts have addressed specific aspects of the Grid Computing infrastructure, such as supporting the federation of independent sites into large scale grids, as the Computational Co op [4] All these projects view Grid Computing infrastructure being deployed as ....

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. ICDCS'98 -- International Conference on Distributed Computing Systems. May 1998.

....ignored the differences between Sarah and Bob s requirements or has gravitated towards Bob s since they are more technically challenging. Moreover, by committing to a particular Grid Computing infrastructure (e.g. Condor pools [10] Globus resources [7] Computational Co ops [4] or Nile farms [1]) both Sarah and Bob must restrict their application to utilize 2 only processors available through such an infrastructure. Bob may be happy to do this, but Sarah might have access to many other processors elsewhere. And, since her application is coarse grained, using whatever processors she can ....

.... to provide comprehensive support for grid computing, a goal that is starting to be pursued commercially by companies such as Entropia [6] Other Grid Computing projects have more focused goals, targeting high throughput applications (as Condor [10] and high energy physics applications (as Nile [1]) for example. Yet other efforts have addressed specific aspects of the Grid Computing infrastructure, such as supporting the federation of independent sites into large scale grids, as the Computational Co op [4] All these projects view Grid Computing infrastructure being deployed as universally ....

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. ICDCS'98 -- International Conference on Distributed Computing Systems. May 1998.

No context found.

Alessandro Amoroso, Keith Marzullo, and Aleta Ricciardi. Wide-area nile: a case study of a wide-area data-parallel application. In Proceedings of the Eighteenth International Conference on Distributed Computing Systems, pages 506-- 515, Amsterdam, Netherlands, 26--29 May 1998.

....and storage resources. These resources are available to process jobs that have been submitted by local users, or legitimate users local to other Nile sites. We focus here on the site local objects and functionality, but local area Nile was designed to adapt easily to a wide area setting [2]. For instance, object groups that must be tightly coupled are not spread across the WAN; conversely, only state information that changes infrequently would be shared at all sites. The Nile design documents describe these criteria in detail [7] CORBA Object Replicated CORBA Object C Object ....

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-Area Nile: A Case Study of a WideArea Data-Parallel Application. Technical Report CS97-566, UC San Diego, 1997.

....and storage resources. These resources are available to process jobs that have been submitted by local users, or legitimate users local to other Nile sites. We focus here on the site local objects and functionality, but local area Nile was designed to adapt easily to a wide area setting [2]. For instance, object groups that must be tightly coupled are not spread across the WAN; conversely, only state information that changes infrequently would be shared at all sites. The Nile design documents describe these criteria in detail [8] Each job is divided into a list of subjobs that are ....

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. UC San Diego TRCS97-566. 1997.

....the application (or some scheduling agent acting on its behalf) decides which clusters to use. There is no global scheduling entity because, due to the non uniform distribution of resources across the sites, good scheduling decisions tend to be application specific in the wide area setting (e.g. [1]) For example, the data a particular application is interested on is likely not to be available on all sites, and its transfer over wide area channels may take a prohibitively long time. Furthermore, there might be other application specific constrains such as avoiding sites deemed as having weak ....

Alessandro Amoroso, Keith Marzullo, and Aleta Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. 18th International Conference on Distributed Computing Systems (ICDCS'98), pp. 506-15, May 1998.

....the application (or some scheduling agent acting on its behalf) decides which clusters to use. There is no global scheduling entity because, due to the non uniform distribution of resources across the sites, good scheduling decisions tend to be application specific in the wide area setting (e.g. Amoroso 98] For example, the data a particular application is interested on is likely not to be available on all sites, and its transfer over wide area channels may take a prohibitively long time. Furthermore, there might be other appli 12 cation specific constrains such as avoiding sites deemed as ....

Alessandro Amoroso, Keith Marzullo, and Aleta Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area Data-Parallel Application. 18th International Con- ference on Distributed Computing Systems (ICDCS'98), pp. 506-15, May 1998.

No context found.

A. Amoroso, K. Marzullo, and A. Ricciardi. Wide-Area Nile: A Case Study of a Wide-Area DataParallel Application. In Proceedings of the 18th International Conference on Distributed Computing Systems (ICDCS), Amsterdam, Netherlands, pages 506-515, May 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC