Abstract
At the advent of a wished (or forced) convergence between High Performance Computing HPC platforms, stand-alone accelerators and virtualized resources from Cloud Computing CC systems, this article unveils the job prediction component of the Evalix project. This framework aims at an improved efficiency of the underlying Resource and Job Management System RJMS within heterogeneous HPC facilities by the automatic evaluation and characterization of the submitted workload. The objective is not only to better adapt the scheduled jobs to the available resource capabilities, but also to reduce the energy costs. For that purpose, we collected the resource consumption of all the jobs executed on a production cluster for a period of three months. Based on the analysis then on the classification of the jobs, we computed a resource consumption model. The objective is to train a set of predictors based on the aforementioned model, that will give the estimated CPU, memory and IO used by the jobs. The analysis of the resource consumption highlighted that different classes of jobs have different kinds of resource needs and the classification of the jobs enabled to characterize several application patterns of the users. We also discovered that several users whose resource usage on the cluster is considered as too low, are responsible for a loss of CPU time on the order of five years over the considered three month period. The predictors, trained from a supervised learning algorithm, were able to correctly classify a large set of data. We evaluated them with three performance indicators that gave an information retrieval rate of 71% to 89% and a probability of accurate prediction between 0.7 and 0.8. The results of this work will be particularly helpful for designing an optimal partitioning of the considered heterogeneous platform, taking into consideration the real application needs and thus leading to energy savings and performance improvements. Moreover, apart from the novelty of the contribution, the accurate classification scheme offers new insights of users behavior of interest for the design of future HPC platforms.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Later on, another monitoring tool named Colmet [11] will be used.
References
Lublin, U., Feitelson, D.: The workload on parallel supercomputers: modeling the characteristics of rigid jobs. J. Parallel Distrib. Comput. 63, 1105–1122 (2001)
Feitelson, D.G.: Workload modeling for performance evaluation. In: Calzarossa, M.C., Tucci, S. (eds.) Performance 2002. LNCS, vol. 2459, pp. 114–141. Springer, Heidelberg (2002). doi:10.1007/3-540-45798-4_6
Feitelson, D.G., Jettee, M.A.: Improved utilization and responsiveness with gang scheduling. In: Feitelson, D.G., Rudolph, L. (eds.) JSSPP 1997. LNCS, vol. 1291, pp. 238–261. Springer, Heidelberg (1997). doi:10.1007/3-540-63574-2_24
Cao, J., Zimmermann, F.: Queue scheduling and advance reservations with cosy. In: Parallel and Distributed Processing Symposium, p. 63 (2004)
Emeras, J., Ruiz, C., Vincent, J.-M., Richard, O.: Analysis of the jobs resource utilization on a production system. In: Desai, N., Cirne, W. (eds.) JSSPP 2013. LNCS, vol. 8429, pp. 1–21. Springer, Heidelberg (2014). doi:10.1007/978-3-662-43779-7_1
Varrette, S., Bouvry, P., Cartiaux, H., Georgatos, F.: Management of an academic HPC cluster: the UL experience. In: Proceedings of the 2014 HPCS Conference (2014)
Capit, N., Costa, G.D., Georgiou, Y., et al.: A batch scheduler with high level components. In: CCGrid, pp. 776–783 (2005)
Wolter, N., McCracken, M.O., Snavely, A., et al.: What’s working in HPC: Investigating HPC user behavior and productivity. CTWatch Q. 2, 9–17 (2006)
Feitelson, D.G., Tsafrir, D., Krakov, D.: Experience with using the parallel workloads archive. J. Parallel Distribut. Comput. 74(10), 2967–2982 (2014)
Feitelson, D.: Parallel workload archive
Linux Kernel: https://www.kernel.org/, Taskstats: https://www.kernel.org/doc/Documentation/accounting/taskstats.txt, Cgroups: https://www.kernel.org/doc/Documentation/cgroups/cgroups.txt
Bailey, D.H.: NAS parallel benchmarks. In: Padua, D. (ed.) Encyclopedia of Parallel Computing. Springer, Heidelberg (2011)
Cortes, C., Vapnik, V.: Support-vector networks. Mach. Learn. 20(3), 273–297 (1995)
Duan, R., Nadeem, F., Wang, J., Zhang, Y., Prodan, R., Fahringer, T.: A hybrid intelligent method for performance modeling and prediction of workflow activities in grids. In: Proceedings of the 2009 CCGRID Conference, pp. 339–347 (2009)
Chang, C.C., Lin, C.J.: Libsvm: a library for support vector machines. ACM Trans. Intell. Syst. Technol. 2(3), 27: 1–27: 27 (2011)
Hsu, C.W., Lin, C.J.: A comparison of methods for multiclass support vector machines. IEEE Trans. Neural Netw. 13(2), 415–425 (2002)
Szollosi, D., Denes, D.L., Firtha, F., Kovacs, Z., Fekete, A.: Comparison of six multiclass classifiers by the use of different classification performance indicators. J. Chemometr. 26(3–4), 76–84 (2012)
Ben-David, A.: Comparison of classification accuracy using cohen’s weighted kappa. Expert Syst. Appl. 34(2), 825–832 (2008)
Provost, F.J., Fawcett, T., et al.: Analysis and visualization of classifier performance: comparison under imprecise class and cost distributions. KDD 97, 43–48 (1997)
Uebersax, J.S.: A generalized kappa coefficient. Educ. Psychol. Meas. 42(1), 181–183 (1982)
Feinstein, A.R., Cicchetti, D.V.: High agreement but low kappa: I. the problems of two paradoxes. J. Clin. Epidemiol. 43(6), 543–549 (1990)
Hand, D., Till, R.: A simple generalisation of the area under the roc curve for multiple class classification problems. Mach. Learn. 45(2), 171–186 (2001)
Bradley, A.P.: The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognit. 30(7), 1145–1159 (1997)
Duan, K., Keerthi, S., Poo, A.N.: Evaluation of simple performance measures for tuning SVM hyperparameters. Neurocomputing 51, 41–59 (2003)
Guyon, I.: A Scaling Law for the Validation-Set Training-Set Size Ratio. AT&T Bell Laboratories (1997)
Matsunaga, A., Fortes, J.A.B.: On the use of machine learning to predict the time and resources consumed by applications. In: CCGrid (2010)
Tsafrir, D., Etsion, Y., Feitelson, D.: Backfilling using system-generated predictions rather than user runtime estimates. IEEE Trans. Parallel Distrib. Syst. 18(6), 789–803 (2007)
Smith, W., Foster, I., Taylor, V.: Predicting application run times using historical information. In: Feitelson, D.G., Rudolph, L. (eds.) JSSPP 1998. LNCS, vol. 1459, pp. 122–142. Springer, Heidelberg (1998). doi:10.1007/BFb0053984
Gibbons, R.: A historical application profiler for use by parallel schedulers. In: Feitelson, D.G., Rudolph, L. (eds.) JSSPP 1997. LNCS, vol. 1291, pp. 58–77. Springer, Heidelberg (1997). doi:10.1007/3-540-63574-2_16
Zhang, J., Figueiredo, R.: Application classification through monitoring and learning of resource consumption patterns. In: IPDPS, April 2006
Acknowledgments
The experiments presented in this paper were carried out using the HPC facility of the University of Luxembourg. Many thanks are also due to all those who participated in collecting and distributing the logs available through the PWA and used in Table 1.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2017 Springer International Publishing AG
About this paper
Cite this paper
Emeras, J., Varrette, S., Guzek, M., Bouvry, P. (2017). Evalix: Classification and Prediction of Job Resource Consumption on HPC Platforms. In: Desai, N., Cirne, W. (eds) Job Scheduling Strategies for Parallel Processing. JSSPP JSSPP 2015 2016. Lecture Notes in Computer Science(), vol 10353. Springer, Cham. https://doi.org/10.1007/978-3-319-61756-5_6
Download citation
DOI: https://doi.org/10.1007/978-3-319-61756-5_6
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-61755-8
Online ISBN: 978-3-319-61756-5
eBook Packages: Computer ScienceComputer Science (R0)