병렬 프로그램 로그 군집화 기반 작업 실행 시간 예측모형 연구
Runtime Prediction Based on Workload-Aware Clustering
0
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
Several fields of science have demanded large-scale workflow support, which requires thousands of CPU cores or more. In order to support such large-scale scientific workflows, high capacity parallel systems such as supercomputers are widely used. In order to increase the utilization of these systems, most schedulers use backfilling policy: Small jobs are moved ahead to fill in holes in the schedule when large jobs do not delay. Since an estimate of the runtime is necessary for backfilling, most parallel systems use user’s estimated runtime. However, it is found to be extremely inaccurate because users overestimate their jobs. Therefore, in this paper, we propose a novel system for the runtime prediction based on workload-aware clustering with the goal of improving prediction performance. The proposed method for runtime prediction of parallel applications consists of three main phases. First, a feature selection based on factor analysis is performed to identify important input features. Then, it performs a clustering analysis of history data based on self-organizing map which is followed by hierarchical clustering for finding the clustering boundaries from the weight vectors. Finally, prediction models are constructed using support vector regression with the clustered workload data. Multiple prediction models for each clustered data pattern can reduce the error rate compared with a single model for the whole data pattern. In the experiments, we use workload logs on parallel systems (i.e., iPSC, LANL-CM5, SDSC-Par95, SDSC-Par96, and CTC-SP2) to evaluate the effectiveness of our approach. Comparing with other techniques, experimental results show that the proposed method improves the accuracy up to 69.08%.
Abstract
Several fields of science have demanded large-scale workflow support, which requires thousands of CPU cores or more. In order to support such large-scale scientific workflows, high capacity parallel systems such as supercomputers are widely used. In order to increase the utilization of these systems, most schedulers use backfilling policy: Small jobs are moved ahead to fill in holes in the schedule when large jobs do not delay. Since an estimate of the runtime is necessary for backfilling, most parallel systems use user’s estimated runtime. However, it is found to be extremely inaccurate because users overestimate their jobs. Therefore, in this paper, we propose a novel system for the runtime prediction based on workload-aware clustering with the goal of improving prediction performance. The proposed method for runtime prediction of parallel applications consists of three main phases. First, a feature selection based on factor analysis is performed to identify important input features. Then, it performs a clustering analysis of history data based on self-organizing map which is followed by hierarchical clustering for finding the clustering boundaries from the weight vectors. Finally, prediction models are constructed using support vector regression with the clustered workload data. Multiple prediction models for each clustered data pattern can reduce the error rate compared with a single model for the whole data pattern. In the experiments, we use workload logs on parallel systems (i.e., iPSC, LANL-CM5, SDSC-Par95, SDSC-Par96, and CTC-SP2) to evaluate the effectiveness of our approach. Comparing with other techniques, experimental results show that the proposed method improves the accuracy up to 69.08%.
참고문헌 (24)
- Agarwal, B., "Text classification using machine learning methods-A survey", Advances in Intelligent Systems and Computing, 236, pp.701~709 (2014)
- Recent developments in document clustering, Technical report, "Andrews, N.O.", Department of Computer Science, ~, pp.TR-07-35 (2007)
- Chandola, V., "Anomaly detection : A survey", ACM Computing Surveys, 41(3), pp.15~57 (2009)
- Deelman, E., "Workflows and e-science : An overview of workflow system features and capabilities", Future Generation Computer Systems, 25(5), pp.528~540 (2009)
- Downey, A.B., "The elusive goal of workload characterization", ACM SIGMETRICS Performance Evaluation Review, 26, pp.14~29 (1999)
- Downey, A. B., "Using queue time predictions for processor allocation", in Proc. of the Workshop on Job Scheduling Strategies for Parallel Processing, 1291, pp.35~57 (1997)
- Drucker, H., "Support vector regression machines. Neural Information Processing Systems", pp.155~161 (1997), MIT Press
- "Parallel workloads archive and standard workload format"[URL]
- Feitelson, D.G., "Job characteristics of a production parallel scientific workload on the NASA Ames iPSC", 860, 1995, pp.949, 337~360
- Gibbons, R., "A historical application profiler for use by parallel schedulers", Lecture Notes on Computer Science, 1297, pp.58~75 (1997)
- Gil, Y., "Examining the challenges of scientific workflows", IEEE Computer, 40(12), pp.24~32 (2007)
- Hardle, W.K., "Applied Multivariate Statistical Analysis" (2012), Springer-Verlag
- Kaufman, L., "Finding Groups in Data : An Introduction to Cluster Analysis" (2009), John Wiley and Sons
- Kohonen, T., "Essentials of the self-organizing map", Neural Networks, 37, pp.52~65 (2013)
- Kohonen, T., "Self-Organizing Map" (2001), Springer-Verlag
- Lifka, D. A., "The ANL", IBM scheduling system, 1995, pp.949, 295~303
- Minh, T.N., "Using historical data to predict application runtimes on backfilling parallel systems", in Proc. of 18th Euromicro Conference on Parallel. Distributed and Network-based Processing, pp.246~252 (2010)
- Mu'alem, A.W., "Utilization, predictability, workloads, and user runtime estimates in scheduling the IBM SP2 with backfilling", IEEE Transactions on Parallel and Distributed Systems, 12(6) (2001)
- "R : A language and environment for statistical computing"[URL]
- Senger, L.J., "An instance-based learning approach for predicting execution times of parallel applications", in Proc. of International Information and Telecommunication Technologies Symposium, pp.9~15 (2005)
- Smith, W., "Predicting application run times with historical information", Journal of Parallel and Distributed Computing, 64(9), pp.1007~1016 (2004)
- Tsafrir, D., "Backfilling using system-generated predictions rather than user runtime estimates", IEEE Transactions on Parallel and Distributed Systems, 18(6) (2007)
- Vapnik, V., "The Nature of Statistical Learning Theory" (1995), Springer-Verlag
- Weinland, D., "A survey of vision-based methods for action representation, segmentation and recognition", Computer Vision and Image Understanding, 115(2), pp.224~241 (2011)