|
DOI: 10.14489/vkit.2018.12.pp.045-056
Казьмина К. П., Антонов А. С.
РАЗРАБОТКА МЕТОДОВ ПРОГНОЗИРОВАНИЯ МАСШТАБИРУЕМОСТИ ПРИЛОЖЕНИЙ НА КОНФИГУРАЦИИ СУПЕРКОМПЬЮТЕРОВ
(с. 45-56)
Аннотация. Рассмотрены различные подходы к прогнозированию масштабируемости приложений и предложено использование результатов запусков приложения на малых конфигурациях вычислительной системы для предсказания значений характеристик исполнения данного приложения на конфигурациях бóльшего размера. Применимость предложенного подхода проверена на суперкомпьютерах МГУ им. М. В. Ломоносова «Ломоносов» и «Ломоносов-2». Результаты получены с использованием оборудования Центра коллективного пользования сверхвысокопроизводительными вычислительными ресурсами МГУ им. М. В. Ломоносова.
Ключевые слова: масштабируемость; прогнозирование; регрессия; предсказание производительности.
Kazmina K. P., Antonov A. S.
DEVELOPMENT OF PREDICTION METHODS FOR PARALLEL APPLICATION SCALABILITY ON SUPERCOMPUTER CONFIGURATIONS
(pp. 45-56)
Abstract. Performance characteristics of a parallel application executed on various supercomputer configurations sometimes have a complex behavior that cannot be described by simple analytic models. In general, parallel application scalability tends to be affected by algorithmic features, input parameter values, machine architecture, availability of hardware resources, etc. For accurate scalability predictions all these factors need to be taken into account. To capture such complexity, empirical data can be used during prediction model construction process. In this paper, we review different approaches to scalability prediction and propose using parallel application executions on several small supercomputer configurations (test executions) to estimate its performance characteristics on larger configurations. We explore three types of prediction techniques to describe dynamic performance characteristics: using extrapolation models; using implicit models; and using explicit models. We describe developed parametric predictive methods of the first two types and compare them based on median prediction errors. We also compare our extrapolation model with existing predictive models for parallel application scalability.Applicability of these approaches is tested on Lomonosov and Lomonosov-2 supercomputers at Lomonosov Moscow State University using executions of Linpack and NAS (NASA Advanced Supercomputing) BT (Block Tridiagonal) benchmarks. Predictions are made for configurations of up to 1024 processes with test executions performed using up to 128 processes.
Keywords: Scalability; Prediction methods; Regression; Performance prediction.
К. П. Казьмина, А. С. Антонов (Московский государственный университет им. М. В. Ломоносова, Москва, Россия) E-mail:
Этот e-mail адрес защищен от спам-ботов, для его просмотра у Вас должен быть включен Javascript
K. P. Kazmina, A. S. Antonov (Lomonosov Moscow State University, Moscow, Russia) E-mail:
Этот e-mail адрес защищен от спам-ботов, для его просмотра у Вас должен быть включен Javascript
(Lomonosov Moscow State University, Moscow, Russia) E-mail:
Этот e-mail адрес защищен от спам-ботов, для его просмотра у Вас должен быть включен Javascript
1. Практика суперкомпьютера «Ломоносов» / Вл. В. Воеводин и др. // Открытые системы. СУБД. 2012. № 7. С. 36 – 39.
2. Antonov A., Teplov A. Generalized Approach to Scalability Analysis of Parallel Applications // Lecture Notes in Computer Science. 2016. V. 10049. P. 291 – 304. doi: 10.1007/978-3-319-49956-7_23
3. A Regression-Based Approach to Scalability Prediction / B. J. Barnes et al. // Proc. of the 22nd Annual Intern. Conf. on Supercomputing. 2008. P. 368 – 377.
4. Comparing Scalability Prediction Strategies on an SMP of CMPs / K. Singh et al. // Euro-Par 2010 – Parallel Pro-cessing / 16th Intern. Euro-Par Conf., Ischia, Italy, August 31 – September 3, 2010. Part I. Lecture Notes in Computer Science. 2010. V. 6271. P. 143 – 155. doi: 10.1007/978-3-642-15277-1_14
5. Exploiting Redundancy and Application Scalability for Cost-Effective, Time-Constrained Execution of HPC Applications on Amazon EC2 / A. Marathe et al. // IEEE Transactions on Parallel and Distributed Systems. 2016. V. 27, No. 9. P. 2574 – 2588.
6. Escobar R., Boppana R. V. Performance Prediction of Parallel Scientific Applications // High-Performance Parallel and Distributed Computing. Poster. 2017.
7. Performance Prediction and Ranking of Supercomputers / T.-Y. Chen et al. // Advances in Computers. 2008. V. 72. P. 135 – 172.
8. Chatzopoulos G., Dragojević A., Guerraoui R. ESTIMA: Extrapolating ScalabiliTy of In-Memory Applications // ACM Transactions on Parallel Computing (TOPC). 2017. V. 4, Is. 2. 10 p.
9. Escobar R., Boppana R. V. Performance Prediction of Parallel Applications Based on Small-Scale Executions // IEEE 23rd Intern. Conf. on High Performance Computing (HiPC). 2016. P. 362 – 371.
10. Methods of Inference and Learning for Performance Modeling of Parallel Applications / B. C. Lee et al. // Proc. of the PPoPP. 2007. P. 249 – 258.
11. An Approach to Performance Prediction for Parallel Applications / E. Ipek et al. // Proc. of European Conf. on Paral-lel. Lecture Notes in Computer Science. 2005. V. 3648. P. 196 – 205.
12. Modeling and Predicting Execution Time of Scientific Workflows in the Grid Using Radial Basis Function Neural Network / F. Nadeem et al. // Cluster Computing. 2017. V. 20, No. 3. P. 2805 – 2819.
13. Energy Prediction of CUDA Application Instances Using Dynamic Regression Models / R. S. Rejitha et al. // Computing. 2017. V. 99, No. 8. P. 765 – 790.
14. FASE: A Framework for Scalable Performance Prediction of HPC Systems and Applications / E. Grobelny et al. // SIMULATION. 2007. V. 83, No. 10. P. 721 – 745.
15. Performance Prediction for Large-Scale Parallel Applications Using Representative Replay / J. Zhai et al. // IEEE Transactions on Computers. 2016. V. 65, No. 7. P. 2184 – 2198.
1. Voevodin V. V. et al. (2012). The practice of the supercomputer "Lomonosov". Otkrytye sistemy. SUBD, (7), pp. 36-39. [in Russian language]
2. Antonov A., Teplov A. (2016). Generalized Approach to Scalability Analysis of Parallel Applications. Lecture Notes in Computer Science, 10049, pp. 291-304. doi: 10.1007/978-3-319-49956-7_23
3. Barnes B. J et al. (2008). A Regression-Based Approach to Scalability Prediction. Proceedings of the 22nd Annual International Conference on Supercomputing, pp. 368-377.
4. Singh K. et al. (2010). Comparing Scalability Predic-tion Strategies on an SMP of CMPs. Euro-Par 2010 – Parallel Processing, 16th International Euro-Par Conference, Ischia, Italy, August 31 – September 3, 2010. Part I. Lecture Notes in Com-puter Science, 6271, pp. 143-155. doi: 10.1007/978-3-642-15277-1_14
5. Marathe A. et al. (2016). Exploiting Redundancy and Application Scalability for Cost-Effective, Time-Constrained Execution of HPC Applications on Amazon EC2. IEEE Trans-actions on Parallel and Distributed Systems, 27(9), pp. 2574- 2588.
6. Escobar R., Boppana R. V. (2017). Performance Prediction of Parallel Scientific Applications. High-Performance Parallel and Distributed Computing. Poster.
7. Chen T.-Y. et al. (2008). Performance Prediction and Ranking of Super-computers. Advances in Computers, 72, pp. 135-172.
8. Chatzopoulos G., Dragojević A., Guerraoui R. (2017). ESTIMA: Extrapolating ScalabiliTy of In-Memory Applications. ACM Transactions on Parallel Computing (TOPC), 4(2).
9. Escobar R., Boppana R. V. (2016). Performance Prediction of Parallel Applications Based on Small-Scale Executions. IEEE 23rd International Conference on High Performance Computing (HiPC), pp. 362-371.
10. Lee B. C. et al. (2007). Methods of Inference and Learning for Performance Modeling of Parallel Applications. Proceedings of the PPoPP, pp. 249-258.
11. Ipek E. et al. (2005). An Approach to Performance Prediction for Parallel Applications. Proceedings of European Conf. on Parallel. Lecture Notes in Computer Science, 3648, pp. 196-205.
12. Nadeem F. et al. (2017). Modeling and Predicting Execution Time of Scientific Workflows in the Grid Using Radial Basis Function Neural Network. Cluster Computing, 20(3), pp. 2805-2819.
13. Rejitha R. S. et al. (2017). Energy Prediction of CUDA Application Instances Using Dynamic Regression Models. Computing, 99(8), pp. 765-790.
14. Grobelny E. et al. (2007). FASE: A Framework for Scalable Performance Prediction of HPC Systems and Applica-tions. SIMULATION, 83(10), pp. 721-745.
15. Zhai J. et al. (2016). Performance Prediction for Large-Scale Parallel Applications Using Representative Replay. IEEE Transactions on Computers, 65(7), pp. 2184-2198.
Статью можно приобрести в электронном виде (PDF формат).
Стоимость статьи 350 руб. (в том числе НДС 18%). После оформления заказа, в течение нескольких дней, на указанный вами e-mail придут счет и квитанция для оплаты в банке.
После поступления денег на счет издательства, вам будет выслан электронный вариант статьи.
Для заказа скопируйте doi статьи:
10.14489/vkit.2018.12.pp.045-056
и заполните форму
Отправляя форму вы даете согласие на обработку персональных данных.
.
This article is available in electronic format (PDF).
The cost of a single article is 350 rubles. (including VAT 18%). After you place an order within a few days, you will receive following documents to your specified e-mail: account on payment and receipt to pay in the bank.
After depositing your payment on our bank account we send you file of the article by e-mail.
To order articles please copy the article doi:
10.14489/vkit.2018.12.pp.045-056
and fill out the form
.
|
Текущий номер
Разработка концепции и создание сайта - ООО «Издательский дом «СПЕКТР»