: Abdul Waheed, Diane T. Rover, Jeffrey Hollingsworth. "Modeling, evaluation and testing of paradyn instrumentation system" Proceedings of the Supercomputing Conference 1996.  Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Automatic Performance Evaluation of Parallel Programs - Espinosa, Margalef, Luque (1998) (4 citations) (Correct)
....of finding which causes match with a specific problem. When all causes of a problem are found, the tool builds a report of suggestions on how to overcome the problems found and presents it to the user programmer of applications. There are other automatic performance analysers, like Paradyn [Waheed 96] that analyse the performance at execution time, therefore they must focus its work in minimising the overhead of introducing monitoring control of the system. For that reason, Paradyn possibly misses problems which fall out from the current instrumentation bounds. In section 2, we describe the ....
: Abdul Waheed, Diane T. Rover, Jeffrey Hollingsworth. "Modeling, evaluation and testing of paradyn instrumentation system" Proceedings of the Supercomputing Conference 1996.
Modeling and Evaluating Design Alternatives for an.. - Waheed, Rover.. (1998) (2 citations) Self-citation (Waheed Rover Hollingsworth) (Correct)
....process and a Paradyn daemon. One of the nodes also executes the main Paradyn process. Paradyn daemons on individual nodes directly forward the instrumentation data to the main process. We select a batch size of 32 samples for the BF policy (based on analysis beyond the scope of this paper; see [30]) For compute intensive applications, the mean network occupancy requirement is arbitrarily set at 200 msec; and for communication intensive applications, 2,000 msec. Four factors of interest in this case are: number of nodes, sampling period, forwarding policy, and application type. Applying the ....
....for the resource contentions according to the scheduling policies used by the operating system. These results indicate that the BF data forwarding policy outperforms the CF policy with respect to both direct overhead and monitoring latency. This is also true for the SMP and MPP architecture (see [30]) Therefore, we consider only the BF policy in the following sections. 3.3.2 SMP Architecture: What is the Effect of Multiple Paradyn Daemons on the Monitoring Latency Simulation results in Fig. 10 show that the monitoring latency increases with the number of nodes. In order to maintain a ....
# A. Waheed, D.T. Rover, and J. Hollingsworth, "Modeling, Evaluation, and Testing of Paradyn Instrumentation System," Proc. Supercomputing `96, Pittsburgh, Pa., Nov. 1996.
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