N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. HPDC, 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....time (or makespan) of a single application executed on behalf of a single user [8, 7, 14, 20, 38, 33] A number of projects pre dating the advent of the Grid have addressed the idea of software as a service on the network. These systems have been traditionally called Network enabled Servers (NES) [13, 32, 22, 15] and are currently in use for many types of applications. These systems usually implement client server architectures and provide users with an RPC style programming model. Users can then easily access (interactively or via programs) computational modules and compute cycles on remote resources. ....

N. Kapadia, J. Forter, and C. Brodley. Predictive Application-Performance Modeling in a Computational Grid Environment. In Processings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC8), 1999.

....predict resource demand as a function of runtime parameters: however, neither uses the predictions for application adaptation. Automated profiling for QoS [1] estimates the CPU utilization of a multimedia stream as a linear function of task rate and task size, for admission control purposes. PUNCH [21] uses machine learning to predict CPU demand as a function of application specific runtime parameters, for load balancing in a grid framework. To the best of our knowledge, Odyssey is the first system to use history based prediction to model resource demand as a function of fidelity in adaptive ....

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive Application-Performance Modeling in a Computational Grid Environment. In Proc. 8th IEEE International Symposium on High Performance Distributed Computing (HPDC '99), pages 47--54, Los Angeles, CA, Aug. 1999.

....rate and average task size in bytes. The technique used linear estimation with forgetting is identical to Recursive Least Squares (Section 6.3.1) The predictions of utilization are then used for QoS admissions control or allocation. More closely related to my work is the PUNCH system [53]. PUNCH uses machine learning techniques to predict the CPU demand of a program run as a function of applicationspecific runtime parameters. This is very similar to history based demand prediction, with more sophisticated (and expensive) learning algorithms than linear estimation. The demand ....

Nirav H. Kapadia, Jose A. B. Fortes, and Carla E. Brodley. Predictive applicationperformance modeling in a computational grid environment. In Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC '99), pages 47--54, Los Angeles, CA, August 1999.

....or static set of resources. However, computational grid environments consists of a collection of dynamic, heterogeneous resources. Our approach seeks to separate the influences of the runtime system from the behavior intrinsic to the application. This is described in more detail in 4.2. Similar to [11], we use information about application input parameters when predicting application performance. However, rather than relying on application performance data from previous runs on the same system, our approach combines the application intrinsic knowledge with run time predictions of resource ....

....for an application, and the appropriate signature can be looked up at runtime. If no signature exists for the specific problem parameters from the current execution, a suitable signature can be chosen or interpolated, for example using the nearest neighbor or interpolation algorithms described in [11]. These are instance based learning algorithms, which predict the output value based on the problem parameters and a set of previously recorded instances. The recorded instances are arranged in the problem parameter space, and the current problem parameters form a query point in this space. The ....

KAPADIA, N., FORTES, J., AND BRODLEY, C. Predictive Application-Performance Modeling in a Computational Grid Environment. In Proceedings of the Eight IEEE Symposium on High- Performance Distributed Computing (Redondo Beach, California, August 1999), pp. 454.

....For example, Gas requires a performance model in order to schedule an application. It is therefore important to determine the minimum accuracy of a performance model for it to be useful for Gas. This will enable us to determine whether automatically generated performance models (such as those in [Kapadia 99] and [Smith 98] are accurate enough to be used by Gas. Bushel of AppLes over Multiple Supercomputers Although we are currently focusing on one supercomputer, the whole motivation behind metacomputing friendly resource schedulers is to empower users to efficiently use multiple resources, ....

Nirav Kapadia, Jos Fortes, and Carla Brodley. Predictive Application -Performance Modeling in a Computational Grid Environment. Eighth IEEE Symposium on High-Performance Distributed Computing, July 1999.

....ne a common set of services and concepts that are necessary for implementing and deploying NES systems on the Computational Grid. This paper also describes current work with scienti c and engineering simulations that are enabled by NES systems in the Grid context. 1 Introduction Several projects [2, 3, 4, 5, 6, 7] aim at providing simple ways (APIs, GUIs) to execute software remotely on the Computational Grid. Typically the software executed by servers consists of scienti c libraries or programs. In what follows, we use the term module to denote that software, whether it is part of a library, a stand alone ....

....nonprogrammable interfaces, typically GUI Web based, to the Grid. These are now coined as scienti c portals or Grid portals . For NES systems, there is work to present the users with what are e ectively portal front ends to access the servers on the backend. A good example is the Punch system [5], which de nes a Web based access framework to specially packaged software library, and has been in active use. Nimrod, NetSolve, and Ninf already provides some form of web based access to their respective infrastructure. More recent work by the Ninf group attempts to unify the portals technology ....

N.H. Kapadia, J.A.B. Forter, and C.E. Brodley. Predictive Application-Performance Modeling in a Computational Grid Environment. In Processings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC8), 1999.

.... This mechanism provides immediate access to computing resources for new NETCARE users, and automatic access to newly added computing resources to the existing user base [12] The resource management subsystem of PUNCH incorporates load balancing and predictive performance modeling mechanisms [14], and provides access to cluster management software across multiple administrative domains. Cluster management systems, e.g. Condor [17] and PBS [2] typically provide a command and or library interface to interact with the system. PUNCH accesses clusters resources via this interface [1] The ....

Kapadia, N. H., Fortes, J. A. B., and Brodley, C. E. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8 th International Symposium on High Performance Distributed Computing (HPDC'99), pages 47--54, Aug. 1999.

....resource consumption as a function of fidelity in order to improve adaptation in mobile applications. We see our predictive mechanism as a service to be used by higher level adaptive systems. We are aware of one other piece of work that tries to learn resource consumption functions: PUNCH [9] is a system for learning the CPU requirements of an application as a function of the input parameters. The objective of PUNCH is to use predictions of CPU usage to decide how and where to execute the application in a distributed computing environment. The Odyssey predictor, on the other hand, ....

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. In Eighth IEEE International Symposium on High Performance Distributed Computing (HPDC), Los Angeles, CA, Aug. 1999.

....This mechanism provides immediate access to computing resources for new NETCARE users, and automatic access to newly added computing resources to the existing user base. The resource management subsystem of PUNCH incorporates load balancing and predictive performance modeling mechanisms [9], and provides access to cluster management software such as Condor [12] This allows the system to provide access to a large number of hardware resources through a single entry point. Currently, the NETCARE infrastructure provides access to 5 dedicated servers at Purdue, and approximately 600 ....

Kapadia, N. H., Fortes, J. A. B., and Brodley, C. E. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8 th International Symposium on High Performance Distributed Computing (HPDC'99), pages 47--54, Aug. 1999.

....best of our knowledge, this is the first piece of work that learns and predicts application resource consumption as a function of fidelity in order to improve adaptation in mobile applications. We are aware of one other piece of work that tries to learn and resource consumption functions: PUNCH [7] is a system for learning the CPU requirements of an application as a function of the input parameters. The objective of PUNCH is to use predictions of CPU usage to decide how and where to execute the application in a distributed computing environment. The Odyssey predictor, on the other hand, ....

Nirav H. Kapadia, Jose A. B. Fortes, and Carla E. Brodley. Predictive application-performance modeling in a computational grid environment. In Eighth IEEE International Symposium on High Performance Distributed Computing (HPDC), Los Angeles, CA, August 1999.

No context found.

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC'99), pages 47--54, Redondo Beach, California, August 1999.

....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, Jos'e A. B. Fortes, and Carla E. Brodley. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC'99), pages 47--54, Redondo Beach, California, August 1999.

....tradeoff at run time, based on specifications of acceptable cost and Purdue University Network Computing Hubs. www.ece.purdue.edu punch. 3 desired performance. Cost figures are provided by administrators. Performance is estimated by way of a predictive application performance modeling system [8] 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 [7] the actual scheduling algorithms can be selected ....

Nirav H. Kapadia, Jos'e A. B. Fortes, and Carla E. Brodley. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC'99), pages 47--54, Redondo Beach, California, August 1999.

.... can specify applicationmanagement metaprograms that automatically select resources based on specified criteria [13] PUNCH has performance modeling system that uses tool specific, run time input in conjunction with machine learning technology to predict the resource usage characteristics of runs [14]. PUNCH also helps distribute runs across multiple machines on the submitting side. For example, Condor places its spool and process image directories on the submitting machine, which requires significant amounts of disk space. PUNCH manages the disk space on these machines by submitting jobs from ....

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC'99), pages 47--54, Redondo Beach, California, August 1999.

.... specify application man agemen t metaprograms that automatically select resources based on speci ed criteria [13] PUNCH has performan ce modelin g system that uses tool speci c, run time inS5 in co nun tion with machin e learn n g techn logy to predict the resource usage characteristics ofrun [14]. PUNCH also helps distribute run across multiple machin es on the submittin g side. For example, Con dor places its spool an d process image directories on the submittin machin e, which requires sign i can t amoun ts of disk space. PUNCH man ages the disk space on these machi n s by submittin ....

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC'99), pages 47--54, Redondo Beach, California, August 1999.

No context found.

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. HPDC, 1999.

No context found.

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. HPDC, 1999.

No context found.

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC '99), pages 47--54, Redondo Beach, CA, Aug. 1999. 26

No context found.

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive Application-Performance Modeling in a Computational Grid Environment. In Proc. 8th IEEE International Symposium on High Performance Distributed Computing (HPDC '99), pages 47--54, Los Angeles, CA, Aug. 1999.

No context found.

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. HPDC, 1999.

No context found.

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performancemodeling in acomputational grid environment. In Eighth IEEE International Symposium on High Performance Distributed Computing (HPDC), Los Angeles, CA, August 1999.

No context found.

N. H. Kapadia, J. A. B. Fortes, and C. E. Brodley. Predictive application-performance modeling in a computational grid environment. In Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC '99), pages 47--54, Redondo Beach, CA, Aug. 1999. 26

No context found.

Kapadia, N.H., J. A. B. Fortes, and C. E. Brodley, "Predictive Application-Performance Modeling in a Computational Grid Environment", 8 th 1EEE lnt'l. Symposium on High Performance Distributed Computing, August 1999.

No context found.

Nirav Kapadia, Jos Fortes, and Carla Brodley. Predictive ApplicationPerformance Modeling in a Computational Grid Environment. Eighth IEEE Symposium on High-Performance Distributed Computing, July 1999.

No context found.

N.H. Kapadia, J.A.B. Fortes, and C.E. Brodley, "Predictive Application-Performance Modeling in a Computational Grid Environment," Proc. Eighth IEEE Int'l Symp. High Performance Distributed Computing, HPDC'99, IEEE Computer Society Press, Los Alamitos, Calif., Aug. 1999, pp. 47-54.

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