N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom. Resource-usage prediction for demand-based network-computing. In Proceedings of the Workshop on Advances in Parallel and Distributed Systems (APADS), pages 372--377, West Lafayette, Indiana, October 1998. IEEE Computer Society.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....for Software Resources. The PUNCH infrastructure is designed to allow administrators to integrate (install) unmodified tools with little effort; access to source or object code is not required. Stand alone tools with batch, text based interactive, and graphical user interfaces are supported [Kapadia et al. 1998]. Tool 6 Delta N. H. Kapadia, J. A. B. Fortes, and M. S. Lundstrom installation is accomplished by way of a specifically designed high level language that describes the behavior of the tool to the PUNCH infrastructure [Kapadia and Fortes 1999a] the description provided by a tool installer is ....

....an application management framework that allows it to focus on achieving the best possible cost and performance tradeoff at run time, based on specifications of acceptable cost and desired performance. Cost figures are provided by administrators. Performance is estimated by a prediction system [Kapadia et al. 1998; Kapadia et al. 1999] that utilizes machine learning techniques in order to estimate run specific resource requirements before a scheduling decision is made. PUNCH utilizes a non preemptive, decentralized, adaptive, sender initiated resource management framework [Kapadia 1999] the actual ....

Kapadia, N. H., Brodley, C. E., Fortes, J. A. B., and Lundstrom, M. S. 1998. Resource-usage prediction for demand-based network-computing. In Proceedings of the Workshop on Advances in Parallel and Distributed Systems (APADS) (West Lafayette, Indiana, October 1998), pp. 372--377. IEEE Computer Society.

No context found.

N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom. Resource-usage prediction for demand-based network-computing. In Proceedings of the Workshop on Advances in Parallel and Distributed Systems (APADS), pages 372--377, West Lafayette, Indiana, October 1998. IEEE Computer Society.

....survey evaluating usefulness of tutorial HPDC 9 Tutorial. Nirav H. Kapadia and Jose A. B. Fortes. 7 6. Additional Information PUNCH can be accessed at www.ece.purdue.edu punch; courtesy accounts are available. Publications describing the PUNCH infrastructure and its applications are available ([1, 2, 3, 4]) Preprints of upcoming articles ( 5, 6] are available on request. For more information, please contact Dr. Nirav Kapadia (kapadia purdue.edu; 765 494 0630) or Prof. Jos e Fortes at (fortes purdue.edu; 765 494 3540) ....

Nirav H. Kapadia, Carla E. Brodley, Jos'e A. B. Fortes, and Mark S. Lundstrom. Resource-usage prediction for demand-based network-computing. In Proceedings of the Workshop on Advances in Parallel and Distributed Systems (APADS), pages 372--377, West Lafayette, Indiana, October 1998. IEEE Computer Society.

....SCION The Parallel Programming Hub D The VLSI Design Hub Internet Intranet Network Desktop Interface The Purdue University Network Computing Hubs (PUNCH) Fig. 2.1. The PUNCH infrastructure. runs are routed to faster machines) A detailed description of PUNCH is available in [20, 23, 21, 22]. Running a typical simulation on PUNCH is a three step process. The first step involves the creation of the input file(s) required for the relevant simulation. In the second step, users define the input parameters (e.g. command line arguments, etc. for the program and start the simulation. ....

Nirav H. Kapadia, Carla E. Brodley, Jos'e A. B. Fortes, and Mark S. Lundstrom. Resource-usage prediction for demand-based network-computing. In Proceedings of the 1998 Workshop on Advances in Parallel and Distributed Systems (APADS), West Lafayette, Indiana, October 1998.

....assists users in the use of the tools and the infrastructure itself. Access to information and resources via PUNCH can be personalized and or restricted according to user specific needs and access rights. Finally, the use of machine learning technology allows PUNCH to predict run times for tools (Kapadia et al. 1998). This information is primarily used for on demand resource management (e.g. longer runs are routed to faster machines) Descriptions of PUNCH are available (Kapadia, Fortes, and Lundstrom 1997; Kapadia and Fortes 1998) 5. PUNCH TRANSACTIONS PUNCH servers process transactions that involve ....

.... use of machine learning technology allows PUNCH to predict run times for tools (Kapadia et al. 1998) This information is primarily used for on demand resource management (e.g. longer runs are routed to faster machines) Descriptions of PUNCH are available (Kapadia, Fortes, and Lundstrom 1997; Kapadia and Fortes 1998). 5. PUNCH TRANSACTIONS PUNCH servers process transactions that involve document serving, directory information, system and process status queries, file manipulation, and tool interface generation. Table 2 shows the data obtained by profiling the user activity on PUNCH over the past twenty eight ....

[Article contains additional citation context not shown here]

N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom, Resource-usage prediction for demandbased network-computing, in: Proceedings of the 1998 Workshop on Advances in Parallel and Distributed Systems (APADS), 1998.

....needs and access rights. Finally, the use of artificial intelligence technology allows PUNCH to predict run times for tools. This information is primarily used for ondemand resource management (e.g. longer runs are routed to faster machines) A detailed description of PUNCH is available in [1, 2, 3]. Running a typical simulation on PUNCH is a three step process. The first step involves the creation of the input file(s) required for the relevant simulation. In the second step, users define the input parameters (e.g. command line arguments, etc. for the program and start the simulation. ....

Nirav H. Kapadia, Carla E. Brodley, Jose A. B. Fortes, and Mark S. Lundstrom. "Resource-Usage Prediction for DemandBased Network-Computing". In Proceedings of the Workshop on Parallel and Distributed Systems (APADS). October 1998. West Lafayette, Indiana. To appear.

.... Consequently, PUNCH utilizes a learning approach to characterize this relationship (for each tool) Locally weighted polynomial regression (e.g. 4,12] is used to predict the resource usage characteristics for each run; a detailed description of the machine learning system is available in [26,27]. 6 Finally, the prediction is used in conjunction with portability information to match the user s request to the underlying network accessible tools and resources. The portability information consists of a list of the available implementations e.g. sequential versus parallel for a given ....

N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom, Resource-usage prediction for demand-based networkcomputing, in: Proceedings of the 1998 Workshop on Advances in Parallel and Distributed Systems (APADS), 1998.

.... Consequently, PUNCH utilizes a learning approach to characterize this relationship (for each tool) Locally weighted polynomial regression (e.g. 4,12] is used to predict the resource usage characteristics for each run; a detailed description of the machine learning system is available in [26,27]. 6 Finally, the prediction is used in conjunction with portability information to match the user s request to the underlying network accessible tools and resources. The portability information consists of a list of the available implementations e.g. sequential versus parallel for a given ....

N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom, Resource usage prediction for demand-based networkcomputing, Technical report TR-ECE 98-9, Department of Electrical and Computer Engineering, Purdue University (1998).

....distributed data [1, 8] This makes locally weighted regression an ideal choice for the given domain. The locally linear model is chosen over the locally quadratic model for two reasons: a) it learns faster (for a locally linear surface) and b) it requires less time to make a prediction [5]. 4.2. Learning Issues The basic LLWR learning algorithm addresses the following issues: a) learning sets of polynomial functions, b) incremental learning, and c) support for irrelevant and unscaled features. Modifications are required to address: a) detection of inadequate feature vectors, b) ....

....local models are generally not used because of the associated computational cost. The subsequent sections present solutions for each of the mentioned problems. Detection of inadequate feature vectors is addressed by storing appropriate metainformation about the instances in the knowledge base [5]. Sensitivity to short term variations without an associated loss in longer term performance is obtained by using a two level knowledge base, which also helps the IBL algorithms scale better. Finally, scalability and noise issues are addressed by: a) not adding all instances to the knowledge base, ....

[Article contains additional citation context not shown here]

N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom. Resource usage prediction for demand-based network-computing. Technical Report TR-ECE 98-9, Department of Electrical and Computer Engineering, Purdue University, 1998.

.... Consequently, a learning approach is used to characterize this relationship (for each tool) Locally weighted polynomial regression (e.g. 3, 11] is used to predict the resourceusage characteristics for each run; a detailed description of the artificial intelligence system is presented in [23, 24]. 6 Finally, the prediction is used in conjunction with portability information to match the user s request to the underlying network accessible tools and resources. The portability information consists of a list of the available implementations e.g. sequential versus parallel for a given ....

N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom. Resource usage prediction for demand-based network-computing. Technical Report TR-ECE 98-9, Department of Electrical and Computer Engineering, Purdue University, 1998.

....and occur at unpredictable times. Consequently, learning algorithms employed for resourceusage prediction must be able to quickly tailor their predictions to short term variations without being unduly affected by them in the longer term. The application performance modeling system for PUNCH [18, 19]: 1) employs locally weighted polynomial regression [2, 9] allowing it to work with unknown feature weights and incomplete noisy information, 2) addresses scalability issues by way of a cache that allows it to exploit the locality of runs [18] and by selectively incorporating information into its ....

....differentiates between short term memory and longterm memory. This paper describes three local learning techniques nearest neighbor, weighted average, and locally weighted regression that can be used to predict resource usage; knowledge representation and management issues are described in [18, 19]. 3. Related Work Existing work aimed at estimating resource usage makes use of cumulative statistical data or analytical expressions to predict run time. Statistical models are typically tool specific, and are constructed from measured execution times of previous runs. Analytical expressions ....

[Article contains additional citation context not shown here]

N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom. Resource usage prediction for demand-based network-computing. Technical Report TR-ECE 98-9, Department of Electrical and Computer Engineering, Purdue University, 1998.

....in emulating an ideal user in terms of its resource management and usage policies. 1. Introduction It is now recognized that the heterogeneous nature of the network computing environment cannot be effectively exploited without some form of adaptive or demand driven resource management (e.g. [10, 11, 12, 14, 18, 27]) A demand driven resource management system can be characterized by its ability to make automatic cost performance tradeoff decisions at run time. Such decisions require that the infrastructure be able to decide how (which implementation e.g. sequential versus parallel) and where (which ....

....average, and locally weighted regression algorithms, respectively. Finally, Section 7 presents the conclusions of this work. 2. Domain Constraints Two sets of problems must be addressed in order to be able to predict tool and run specific resource usage in a computational grid environment [18]. The first set consists of issues that are a consequence of the diversity of the tools executed on the computational grid, Eighth IEEE International Symposium on High Performance Distributed Computing, August 1999. 48 whereas the second set includes issues that arise due to the dynamic nature of ....

[Article contains additional citation context not shown here]

N. H. Kapadia, C. E. Brodley, J. A. B. Fortes, and M. S. Lundstrom. Resource-usage prediction for demand-based network-computing. In Proceedings of the 1998 Workshop on Advances in Parallel and Distributed Systems (APADS), West Lafayette, Indiana, October 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