Gary Shao, Rich Wolski, and Fran Berman. Modeling the cost of redistribution in scheduling. In Eighth SIAM Conference on Parallel Processing for Scientific Computation, March 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to achieve satisfactory performance in every case. In other words, static performance tuning remains inadequate for applications which exhibit non predictable execution dynamics, and for computing platforms characterized by a high dynamicity. For these dynamic systems and environments, adaptive [9 13] performance tuning strategies appear necessary. Traditionally, adaptive distributed programs are able to react in response to changes by altering the system configuration, so that the action most appropriate to preserve (or even to increase) the efficiency of the application is taken when a ....

G. Shao, R. Wolski, F. Berman, Modeling the cost of redistribution in scheduling, in: Proceedings of the Eighth SIAM Conference on Parallel Processing for Scientific Computing, 1997.

....and distributed simulation research in the area of adaptive performance tuning strategies. A (somewhat) recent survey browses through these achievements [2] Not only was the PADS community attracted by the new performance issues , the whole parallel and distributed community was concerned [3, 4, 5, 6, 7]. Within PADS, early experiments tried to map several logical (simulation) processes (LPs) onto a single processor within a distributed execution environment. Automated (static) load balancing was studied in [8, 9] while dynamic strategies (mostly based on migration policies for LPs [10] were ....

G. Shao, R. Wolski, and F. Berman, "Modeling the Cost of Redistribution in Scheduling", in Proceedings of the 8th SIAM Conference on Parallel Processing for Scientific Computing, 1997.

....is based on both of these approaches. Others that take a similar approach include [PB95, GAL96] Unfortunately, if even one phase is unanalyzable, the compiler cannot infer an effective global distribution. Approaches employing a run time system, such as ALEXI [Who91] CHAOS [HMS 95] AppLeS [SWB97] and Adapt [LA96] can find an efficient local data distribution even in cases where workload and communication characteristics of a program change at run time. However, a pure run time system does not necessarily know when phases or phase cycles begin or end, and hence might be unable to ....

Gary Shao, Rich Wolski, and Fran Berman. Modeling the cost of redistribution in scheduling. In Eighth SIAM Conference on Parallel Processing for Scientific Computation, March 1997.

.... models impact the rescheduling process ffl Can we use application benchmarks as well as NWS and system benchmarks to drive the decision to reschedule We have done some preliminary work in developing an application specific rescheduling model for a rescheduling Jacobi 2D AppLeS prototype [SWB97] however much more needs to be done. We are currently exploring how best to make the re scheduling decision for individual AppLeS and when. 4.5 The Bushel of AppLeS Problem The previous projects are focused on the improvement of the performance of individual AppLeS agents. However we envision ....

G. Shao, R. Wolski, and F. Berman. Modeling the cost of redistribution in scheduling. Proceedings of the Eighth SIAM Conference on Parallel Processing for Scientific Computing, 1997.

No context found.

Gary Shao, Rich Wolski, and Fran Berman. Modeling the cost of redistribution in scheduling. In Eighth SIAM Conference on Parallel Processing for Scientific Computation, March 1997.

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