ABSTRACT
As the most widely used parallel job scheduling strategy in production schedulers, EASY has achieved great success, not only because it can balance fairness and performance, but also because it is universally applicable to most HPC systems. However, unfairness still exists in EASY. For real workloads used in this work, our simulation shows that a blocked job can be delayed by later jobs for more than 90 hours. In addition, EASY cannot directly employ parallel job runtime prediction techniques, because this would lead to a serious situation called reservation violation.
In this paper, we aim at guaranteeing strict fairness (no job is delayed by any jobs of lower priority) while achieving attractive performance, and employing prediction without causing reservation violation in parallel job scheduling. We propose two novel strategies, shadow load preemption (SLP) and venture backfilling (VB), which are together integrated into EASY to construct a preemptive venture EASY backfilling (PV-EASY) strategy. Experimental results on three workloads of real HPC systems demonstrate that: First, PV-EASY guarantees strict fairness, in addition to avoiding reservation violation when employing job runtime prediction techniques in scheduling; Second, PV-EASY achieves the same performance as EASY, and outperforms prediction employed EASY; Third, the preemption in PV-EASY is not resource costly and simple enough to be implemented in all HPC systems where EASY works. These advantages make PV-EASY more attractive than EASY in parallel job scheduling, both from academic and industry perspectives.
- }}Lifka, D. A., The ANL/IBM SP scheduling system. In 1st Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 1995. Google ScholarDigital Library
- }}Feitelson, D. G., Experimental analysis of the root causes of performance evaluation results: a backfilling case study. IEEE Transactions on Parallel and Distributed Systems, 2005: p. 175--182. Google ScholarDigital Library
- }}Mu'Alem, A. W. and Feitelson, D. G., Utilization, predictability, workloads, and user runtime estimates in scheduling the IBM SP 2 with backfilling. IEEE Transactions on Parallel and Distributed Systems, 2001. 12(6): p. 529--543. Google ScholarDigital Library
- }}Karger, D., Stein, C., and Wein, J., Scheduling algorithms. CRC Handbook of Computer Science, 1997.Google Scholar
- }}Sgall, J. On-line scheduling - a survey. In A. Fiat and G. Woeginger, editors, On-Line Algorithms: The State of the Art, Lecture Notes in Computer Science, pages 196--231. Springer-Verlag, 1998. Google ScholarDigital Library
- }}Majumdar, S., Eager, D. L., and Bunt, R. B., Scheduling in multiprogrammed parallel systems. ACM SIGMETRICS Performance Evaluation Review, 1988. 16(1): p. 104--113. Google ScholarDigital Library
- }}Sevcik, K. C., Application scheduling and processor allocation in multiprogrammed parallel processing systems. Journal of Performance Evaluation, 1994. 19: p. 107--140 Google ScholarDigital Library
- }}AuYoung, A., Vahdat A., and Snoeren, A. C., Evaluating the Impact of Inaccurate Information in Utility-Based Scheduling. In Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis (SC), 2009. Google ScholarDigital Library
- }}Etsion, Y. and Tsafrir, D., A Short Survey of Commercial Cluster Batch Schedulers. Technical Report 2005--13, The Hebrew University of Jerusalem, May 2005.Google Scholar
- }}Chiang, S. H., Arpaci-Dusseau, A., and Vernon, M. K., The impact of more accurate requested runtimes on production job scheduling performance. In 8th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 2002. Google ScholarDigital Library
- }}Srinivasan, S., Kettimuthu, R., Subramani, V., and Sadayappan, P., Selective reservation strategies for backfill job scheduling. In 8th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 2002. Google ScholarDigital Library
- }}Ward, W. A., Mahood, C. L. and West, J. E., Scheduling jobs on parallel systems using a relaxed backfill strategy. In 8th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 2002. Google ScholarDigital Library
- }}Shmueli, E. and Feitelson, D. G., Backfilling with lookahead to optimize the packing of parallel jobs. Journal of Parallel and Distributed Computing, 2005. 65(9): p. 1090--1107. Google ScholarDigital Library
- }}Jones, J. P. and Nitzberg, B., Scheduling for parallel supercomputing: a historical perspective of achievable utilization. In 5th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 1999. Google ScholarDigital Library
- }}Talby, D. and Feitelson, D. G., Supporting Priorities and Improving Utilization of the IBM SP Scheduler Using Slack-Based Backfilling. In Proceedings of the 13th International Symposium on Parallel Processing (IPPS), 1999. Google ScholarDigital Library
- }}Tsafrir, D., Etsion, Y. and Feitelson, D. G., Backfilling using system-generated predictions rather than user runtime estimates. IEEE Transactions on Parallel and Distributed Systems, 2007. 18(6): p. 789. Google ScholarDigital Library
- }}Thebe, O., Bunde, D. P. and Leung. V. J., Scheduling Restartable Jobs with Short Test Runs. In 14th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 2009. Google ScholarDigital Library
- }}Guim, F., Rodero, I. and Corbalan, J., The resource usage aware backfilling. In 14th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 2009. Google ScholarDigital Library
- }}Kurian, R., Balaji, P. and Sadayappan, P., Opportune job shredding: An effective approach for scheduling parameter sweep applications. In Los Alamos Computer Science Institute Symposium, New Mexico, 2003.Google Scholar
- }}Sabin, G., et al., Scheduling of parallel jobs in a heterogeneous multi-site environment. In 9th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 2003.Google ScholarCross Ref
- }}Shmueli, E. and Feitelson, D. G., On simulation and design of parallel-systems schedulers: are we doing the right thing?. IEEE Transactions on Parallel and Distributed Systems, 2009. 20(7): p. 983--996 Google ScholarDigital Library
- }}Raz, D., Levy, H. and Avi-Itzhak, B., A resource-allocation queueing fairness measure. ACM SIGMETRICS Performance Evaluation Review, 2004. 32(1): p. 130--141. Google ScholarDigital Library
- }}Avi-Itzhak, B., Levy, H. and Raz, D., Quantifying fairness in queueing systems: Principles and applications, in the Engineering and Informational Sciences, v. 22 n. 4, p. 495--517, October 2008. Google ScholarDigital Library
- }}Isard, M., et al., Quincy: Fair Scheduling for Distributed Computing Clusters. In ACM SIGOPS 22nd symposium on Operating systems principles (SOSP), 2009. Google ScholarDigital Library
- }}Mann, L., Queue culture: The waiting line as a social system. The American Journal of Sociology, 1969. 75(3): p. 340--354.Google Scholar
- }}Larson, R. C., Perspectives on queues: social justice and the psychology of queueing. Operations Research, 1987. 35(6): p. 895--905.Google Scholar
- }}Sabin, G. and Kochhar, G., Job Fairness in Non-Preemptive Job Scheduling. In Proceedings of the 2004 International Conference on Parallel Processing (ICPP), 2004 Google ScholarDigital Library
- }}Avi-Itzhak, B., Brosh, E. and Levy, H., SQF: A slowdown queueing fairness measure. Performance Evaluation, 2007. 64(9--12): p. 1121--1136 Google ScholarDigital Library
- }}Ngubiri, J. and van Vliet, M., Characteristics of fairness metrics and their effect on perceived scheduler effectiveness. 2007, Technical Report, Radboud University Nijmegen.Google Scholar
- }}Lee, C. B. and Snavely, A., On the user-scheduler dialogue: Studies of user-provided runtime estimates and utility functions. International Journal of High Performance Computing Applications, 2006. 20(4): p. 495. Google ScholarDigital Library
- }}Lee, C. B., et al., Are user runtime estimates inherently inaccurate?. In 10th Workshop on Job Scheduling Strategies for Parallel Processing (JSSPP), 2005. Google ScholarDigital Library
- }}Tang, W., Lan, Z., Desai, N. and Buettner, D., Fault-Aware, Utility-Based Job Scheduling on Blue Gene/P Systems. In 2009 IEEE International Conference on Cluster Computing (Cluster), 2009.Google ScholarCross Ref
- }}Susukita, R., et al. Performance prediction of large-scale parallell system and application using macro-level simulation, in Proceedings of the 2008 ACM/IEEE conference on Supercomputing (SC), 2008. Google ScholarDigital Library
- }}Kapadia, N. H., Fortes, J. and Brodley, C. E., Predictive application-performance modeling in a computational grid environment. In 8th IEEE Int'l Symp. on High Performance Distributed Computing (HPDC), p. 6, Aug 1999. Google ScholarDigital Library
- }}Krishnaswamy, S., Loke, S. W. and Zaslavsky, A., Estimating computation times of data-intensive applications. IEEE Distributed Systems Online, 2004. 5(4).Google ScholarCross Ref
- }}Lee, C. B. and Snavely, A. E., Precise and realistic utility functions for user-centric performance analysis of schedulers, In 16th International Symposium on High Performance Distributed Computing (HPDC), 2007 Google ScholarDigital Library
- }}Perkovic, D. and Keleher, P. J., Randomization, speculation, and adaptation in batch schedulers. in Proceedings of the 2000 ACM/IEEE conference on Supercomputing (SC). 2000. Google ScholarDigital Library
- }}Zotkin, D. and Keleher, P. J., Job-Length Estimation and Performance in Backfilling Schedulers. in Proceedings of the 8th IEEE International Symposium on High Performance Distributed Computing (HPDC). 1999 Google ScholarDigital Library
- }}Tsafrir, D., Feitelson, D. G.: The dynamics of backfilling: solving the mystery of why increased inaccuracy may help. In: IEEE International Symposium on Workload Characterization, pp. 131--141 (2006)Google ScholarCross Ref
- }}Nadeem, F. and Fahringer, T., Predicting the execution time of grid workflow applications through local learning. In Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis (SC), 2009 Google ScholarDigital Library
- }}Parallel Workloads Archive, http://www.cs.huji.ac.il/labs/parallel/workload/.Google Scholar
- }}Yero, E. and Henriques, M., Contention-sensitive static performance prediction for parallel distributed applications. Performance Evaluation, 2006. 63(4--5): p. 265--277. Google ScholarDigital Library
- }}Jackson, D., Maui/Moab default configuration. with CTO of Cluster Resources, 2006Google Scholar
Index Terms
- PV-EASY: a strict fairness guaranteed and prediction enabled scheduler in parallel job scheduling
Recommendations
Preemptive Hadoop Jobs Scheduling under a Deadline
SKG '12: Proceedings of the 2012 Eighth International Conference on Semantics, Knowledge and GridsMapReduce has become the dominant programming model in a cloud-based data processing environment, such as Hadoop. First In First Out (FIFO) is the default job scheduling policy of Hadoop, but it cannot guarantee that the job will be completed by a ...
Walltime Prediction and Its Impact on Job Scheduling Performance and Predictability
Job Scheduling Strategies for Parallel ProcessingAbstractFor more than two decades researchers have been analyzing the impact of inaccurate job walltime (runtime) estimates on the performance of job scheduling algorithms, especially the backfilling. In this paper, we extend these existing works by ...
Multi-Level Job Flow Cyclic Scheduling in Grid Virtual Organizations
Distributed environments with the decoupling of users from resource providers are generally termed as utility Grids. The paper focuses on the problems of efficient job flow distribution and scheduling in virtual organizations (VOs) of utility Grids ...
Comments