Evaluación del uso de redes neuronales artificiales para predecir el rendimiento de aplicaciones distribuidas independientemente de la plataforma

Jesus Flores-Contreras; Hector Duran-Limon; Efren Mezura-Montes

doi:10.29105/vtga3.3-1119

Autores/as

Jesus Flores-Contreras Universidad Autónoma de Nuevo León
Hector Duran-Limon University of Guadalajara
Efren Mezura-Montes Universidad Veracruzana

DOI:

https://doi.org/10.29105/vtga3.3-1119

Palabras clave:

Sistemas distribuidos, Plataforma independiente, Predicción del tiempo de ejecución, Técnica de aprendizaje automático, Aplicaciones de larga duración

Resumen

Un sistema distribuido es un conjunto de sistemas informáticos independientes interconectados por una red, que trabajan en cooperación y se comportan como un único sistema creando una plataforma subyacente para distintos tipos de aplicaciones. Estas plataformas suelen utilizarse para ejecutar aplicaciones de larga duración que requieren muchos recursos informáticos, como potencia de procesamiento de la CPU, memoria y ancho de banda de la red. En estos sistemas, es importante gestionar los recursos disponibles de forma eficiente para mejorar el rendimiento general del sistema. Saber cómo se va a comportar el tiempo de ejecución de una aplicación puede mejorar mucho el rendimiento de un sistema, ya que esta información permite asignar de forma eficiente los recursos disponibles. En este artículo, presentamos una evaluación de la idoneidad de las redes neuronales artificiales para lograr un enfoque independiente de la plataforma para la predicción del tiempo de ejecución de aplicaciones distribuidas que se ejecutan en sistemas multinúcleo. Realizamos nuestra evaluación con tres aplicaciones paralelas de larga ejecución, a saber, el modelo de Investigación y Predicción Meteorológica (WRF), Octopus y miniFE. Nuestros resultados indican que las redes neuronales son capaces de producir resultados precisos cuando predicen el tiempo de ejecución de la aplicación en la misma plataforma, pero su precisión disminuye al cambiar de plataforma.

Descargas

Los datos de descargas todavía no están disponibles.

Citas

[AL-Dlaeen and Alashqur, 2014] AL-Dlaeen, D. and Alashqur, A. (2014). Using decision tree classification to assist in the prediction of alzheimer’s disease. In Computer Science and Information Technology (CSIT), 2014 6th International Conference on, pages 122–126.

[Carrington et al., 2006] Carrington, L., Snavely, A., and Wolter, N. (2006). A performance prediction framework for scientific applications. Future Gener. Comput. Syst., 22(3):336–346.

[Developers, 2000] Developers, O. T. (2000). Octopus. Accesed : 27/11/2014.

[Dodonov and de Mello, 2010] Dodonov, E. and de Mello, R. F. (2010). A novel approach for distributed application scheduling based on pre- diction of communication events. Future Generation Computer Systems, 26(5):740 – 752.

[Dong et al., 2012] Dong, F., Luo, J., Song, A., Cao, J., and Shen, J. (2012). An effective data aggregation based adaptive long term cpu load prediction mechanism on computational grid. Future Gener. Comput. Syst., 28(7):1030–1044.

[Gianni et al., 2010] Gianni, D., Iazeolla, G., and D’Ambrogio, A. (2010). A methodology to predict the performance of distributed simulations. In Principles of Advanced and Distributed Simulation (PADS), 2010 IEEE Workshop on, pages 1 –9.

[Hayashida et al., 2014] Hayashida, T., Nishizaki, I., Sekizaki, S., and Nishida, M. (2014). Structural optimization of neural networks and train- ing data selection method for prediction. In Computational Intelligence and Applications (IWCIA), 2014 IEEE 7th International Workshop on, pages 171–176.

[Hu et al., 2012] Hu, L., Che, X.-L., and Zheng, S.-Q. (2012). Online system for grid resource monitoring and machine learning-based prediction. Parallel and Distributed Systems, IEEE Transactions on, 23(1):134 –145.

[Ipek et al., 2005] Ipek, Engin, de Supinski, Brunis R., Schulz, Martin, and McKee, Sally A. An Approach to Performance Prediction for Parallel Applications. Cunha Jose, and Medeiro Pedros, editors, Euro-Par, pages 196-205. Springer 2005.

[Kundu et al., 2010] Kundu, S., Rangaswami, R., Dutta, K., and Zhao, M. (2010). Application performance modeling in a virtualized environment. In High Performance Computer Architecture (HPCA), 2010 IEEE 16th International Symposium on, pages 1 –10.

[Li et al., 2009] Li, B., Peng, L., and Ramadass, B. (2009). Accurate and efficient processor performance prediction via regression tree based modeling. J. Syst. Archit., 55:457–467.

[Li et al., 2014] Li, J., Ji, X., Jia, Y., Zhu, B., Wang, G., Li, Z., and Liu, X. (2014). Hard drive failure prediction using classification and regression trees. In Dependable Systems and Networks (DSN), 2014 44th Annual IEEE/IFIP International Conference on, pages 383394.

[Matsunaga and Fortes, 2010] Matsunaga, A. and Fortes, J. (2010). On the use of machine learning to predict the time and resources consumed by applications. In Cluster, Cloud and Grid Computing (CCGrid), 2010 10th IEEE/ACM International Conference on, pages 495–504.

[Mesoscale and of NCAR, 2007] Mesoscale and of NCAR, M. M. M. D. (2007). Weather research and forecasting model. Accesed : 27/11/2014.

[Michael A. Heroux, 2013] Michael A. Heroux, S. N. L. (2013). Finite element mini application minife. Accesed : 27/11/2014.

[Oliner et al., 2011] Oliner, A., Ganapathi, A., and Xu, W. (2011). Advances and challenges in log analysis. Queue, 9(12):30:30–30:40.

[Oyamada et al., 2008] Oyamada, M. S., Zschornack, F., and Wagner, F. R. (2008). Applying neural networks to performance estimation of embedded software. Journal of Systems Architecture, 54(12):224 – 240.

[Qian et al., 2008] Qian, Z., Zeng, M., Qi, D., and Xu, K. (2008). A dynamic scheduling algorithm for distributed kahn process networks in a cluster environment. In Computational Intelligence and Industrial Application, 2008. PACIIA ’08. Pacific-Asia Workshop on, volume 2, pages 36 –42.

[Raychaudhuri, 2008] Raychaudhuri, S. (2008). Introduction to monte carlo simulation. In Simulation Conference, 2008. WSC 2008. Winter, pages 91–100.

[Raza and Baharudin, 2012] Raza, M. and Baharudin, Z. (2012). A review on short term load forecasting using hybrid neural network techniques. In Power and Energy (PECon), 2012 IEEE International Conference on, pages 846–851.

[Sanjay and Vadhiyar, 2008] Sanjay, H. A. and Vadhiyar, S. (2008). Perfor- mance modeling of parallel applications for grid scheduling. J. Parallel Distrib. Comput., 68(8):1135–1145.

[Sarangi and Bhattacharya, 2005] Sarangi, A. and Bhattacharya, A. (2005). Comparison of artificial neural network and regression models for sedi- ment loss prediction from banha watershed in india. Agricultural Water Management, 78(3):195 – 208.

[Sharkawi et al., 2012] Sharkawi, S., DeSota, D., Panda, R., Stevens, S., Taylor, V., and Wu, X. (2012). Swapp: A framework for performance projections of hpc applications using benchmarks. In Parallel and Distributed Processing Symposium Workshops PhD Forum (IPDPSW), 2012 IEEE 26th International, pages 1722–1731.

[Shimizu et al., 2009] Shimizu, S., Rangaswami, R., Duran-Limon, H. A., and Corona-Perez, M. (2009). Platform-independent modeling and prediction of application resource usage characteristics. Journal of Systems and Software, 82(12):2117 – 2127.

[Tsafrir et al., 2007] Tsafrir, D., Etsion, Y., and Feitelson, D. G. (2007). Backfilling using system-generated predictions rather than user runtime estimates. IEEE Trans. Parallel Distrib. Syst., 18(6):789–803.

[Upadhyaya et al., 2013] Upadhyaya, S., Farahmand, K., and Baker- Demaray, T. (2013). Comparison of NN and LR classifiers in the context of screening native american elders with diabetes. Expert Systems with Applications, 40(15):5830 – 5838.

[Yang et al., 2005] Yang, L. T., Ma, X., and Mueller, F. (2005). Cross- platform performance prediction of parallel applications using partial execution. In Proceedings of the 2005 ACM/IEEE conference on Super- computing, SC ’05, pages 40–, Washington, DC, USA. IEEE Computer Society.