Feitelson DG. Memory usage in the LANL CM-5 workload. Proceedings of the IPPS Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP) (Lecture Notes in Computer Science, vol. 1291). Springer: Berlin, 1997; 78--94.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....coscheduling the parallel jobs becomes increasingly dicult. Athough the proposed scheme achieves subpar performance with a large MPL, it is expected perform well under normal circumstances in which the MPL is usually kept small in order to minimize the overhead, especially from paging activity [12]. Fig. 4 plots the average job wait time under batch scheduling (without back lling) and proposed coscheduling technique with varying time slice length and MPL. In this experiment, we submitted 100 NASPBs to the system at once and measured the wait (or queueing) time of each job. The workload ....

D. G. Feitelson. Memory Usage in the LANL CM-5 Workload. In Proc. IPPS'97 Workshop on Job Scheduling St rategies for Parallel Processing, pages 78-94, 1997.

....job s memory requirement ( 16, 19] 3. NCSA Workload Characterization This section provides an overview of the O2K batch workloads used to evaluate the scheduling policies in Section 4, with particular attention to the job characteristics that differ from system workloads reported previously in [6, 9, 4, 5, 22]. Table 4 summarizes the one month NCSA O2K workloads obtained from system logs during October 1999 through March 2000. The processor demand is the product of a job s requested number of processors and actual runtime, summed over all jobs (in a given class) and expressed as a fraction of the ....

....work (e.g. 6] The scheduling policy evaluations in this paper are performed for all jobs that arrive during the month; thus, the arrival rate behavior is provided simply to characterize the workload that is used to evaluate the policies. Similar to the workloads previously reported in [6, 9, 4, 5, 22], a large fraction of the jobs are serial and many jobs request power of 2 processors. In some cases, a higher fraction of the jobs submitted during weekdaylow periods, or on weekends, request the maximum possible (i.e. 64) processors, as in [6] During a typical weekday peak period, a ....

[Article contains additional citation context not shown here]

D. G. Feitelson. Memory usage in the LANL CM-5 workload. Proc. 3rd Workshop on Job Scheduling Strategies for Parallel Processing, Geneva, Apr. 1997, pp. 78-94. Lecture Notes in Comp. Sci. Vol. 1291, Springer-Verlag.

....on the 1024 processor CM 5 at Los Alamos National Laboratory. The conclusion was that memory requirements are also hyperexponentially distributed with strong discrete components. Additionally, there was a weak correlation between the CPU and memory requirements for the job stream studied [4]. We generalize these results to a # resource workload as follows. The multiple resource requirements for a job in the job stream are described by two parameters. The ### resource requirement for job #, # # # , is drawn from a hyperexponential distribution with mean # # . Additionally, the ....

D. G. Feitelson. Memory usage in the lanl cm-5 workload. In D. Feitelson and L. Rudolph, editors, Job Scheduling Strategies for Parallel Processing, volume 1291, pages 78--94. Springer-Verlag, New York, 1997. LNCS.

.... knowledge to evaluate server performance and caching schemes [5, 18, 6] Other examples include studies of process arrivals and runtimes [12, 37] le systems [36] and video streams [48] In the area of parallel systems, descriptive studies of workloads have only started to appear in recent years [29, 76, 58, 27, 14]. There are also some attempts at modeling [10, 28, 21, 41, 23, 54, 15] and on line characterization [34] But where does the data come from There are two main options: use data that is available anyway, or collect data speci cally for the workload model. The latter can be done in two ways: ....

D. G. Feitelson, \Memory usage in the LANL CM-5 workload". In Job Scheduling Strategies for Parallel Processing, D. G. Feitelson and L. Rudolph (eds.), pp. 78-94, Springer Verlag, 1997. Lect. Notes Comput. Sci. vol. 1291.

.... then, several other datasets have been studied, typically emphasizing job sizes and runtimes [27, 14, 15, 6, 2, 1, 18] However, some new attributes have also been considered, such as speedup characteristics, memory usage, user estimates of runtime, and the probability that a job be cancelled [20, 10, 19, 2]. 7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 10 100 1000 10000 100000 runtime [s] 1 2 nodes 2 4 nodes 4 16 nodes 16 400 nodes all jobs Fig. 3. The cumulative distribution functions of runtimes of jobs with di erent sizes, from the SDSC Paragon. 3.3 Some Answers and More ....

....the distributions of runtimes for small and large jobs do tend to be di erent, with large jobs often having longer runtimes [7] Figure 3) This favors the memory bound or xed time scaling models, and contradicts the xed work model. There is also some evidence that larger jobs use more memory [10]. Thus, within a sin 8 gle machine, parallelism is in general not used for speedup but for solving larger problems. Direct evidence regarding user runtime estimates is available in the logs of machines that use back lling. This data reveals that users typically overestimate job runtime by a ....

D. G. Feitelson, \Memory usage in the LANL CM-5 workload". In Job Scheduling Strategies for Parallel Processing, D. G. Feitelson and L. Rudolph (eds.), pp. 78-94, Springer Verlag, 1997. Lect. Notes Comput. Sci. vol. 1291.

No context found.

Feitelson DG. Memory usage in the LANL CM-5 workload. Proceedings of the IPPS Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP) (Lecture Notes in Computer Science, vol. 1291). Springer: Berlin, 1997; 78--94.

No context found.

Dror G. Feitelson. Memory usage in the LANL CM-5 workload. In for Parallel Processing, pages 78--94. Springer Verlag, 1997. Lect. Notes Comput. Sci. vol. 1291.

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