Predicting I/O Performance in HPC Using Artificial Neural Networks


  • Jan Fabian Schmid Universität Hamburg, Hamburg
  • Julian M. Kunkel German Climate Computing Center (DKRZ), Hamburg



The prediction of file access times is an important part for the modeling of supercomputer's storage systems. These models can be used to develop analysis tools which support the users to integrate efficient I/O behavior.
In this paper, we analyze and predict the access times of a Lustre file system from the client perspective. Therefore, we measure file access times in various test series and developed different models for predicting access times.  The evaluation shows that in models utilizing artificial neural networks the average prediciton error is about 30% smaller than in linear models. A phenomenon in the distribution of file access times is of particular interest: File accesses with identical parameters show several typical access times.The typical access times usually differ by orders of magnitude and can be explained with a different processing of the file accesses in the storage system - an alternative I/O path. We investigate a method to automatically determine the alternative I/O path and quantify the significance of knowledge about the internal processing. It is shown that the prediction error is improved significantly with this approach.


John S. Bucy, Jiri Schindler, Steven W. Schlosser, and Gregory R. Ganger. The DiskSim Simulation Environment Version 4.0 Reference Manual, 2008.

Jason Cope, Ning Liu, Sam Lang, Phil Carns, Chris Carothers, and Robert Ross. CODES: Enabling co-design of Multilayer Exascale Storage Architectures. In Proceedings of the Workshop on Emerging Supercomputing Technologies, volume 2011, 2011.

Adam Crume and Carlos Maltzahn. Latent Frequency Synthesis for Behavioral Hard Disk Drive Access Time Models.

Adam Crume, Carlos Maltzahn, Lee Ward, Thomas Kroeger, Matthew Curry, and Ron Oldfield. Fourier-assisted Machine Learning of Hard Disk Drive Access Time Models. In Proceedings of the 8th Parallel Data Storage Workshop, PDSW ’13, pages 45–51, New York,NY, USA, 2013. ACM.

G. Cybenko. Approximation by Superpositions of a Sigmoidal Function. Mathematics of Control, Signals, and Systems, 2:303–314, 1989.

Chengjun Dai, Guiquan Liu, Lei Zhang, and Enhong Chen. Storage Device Performance Prediction with Hybrid Regression Models. In Proceedings of the 2012 13th International Conference on Parallel and Distributed Computing, Applications and Technologies, PDCAT’12, pages 556–559, Washington, DC, USA, 2012. IEEE Computer Society.

Julian Kunkel, Michaela Zimmer, and Eugen Betke. Using Machine Learning to Predict the Performance of Non-Contiguous I/O, 07 2015.

Julian M Kunkel. Simulating Parallel Programs on Application and System Level. Computer Science-Research and Development, 28(2-3):167–174, 2013.

Yonggang Liu, Renato Figueiredo, Dulcardo Clavijo, Yiqi Xu, and Ming Zhao. Towards simulation of parallel file system scheduling algorithms with PFSsim. In Proceedings of the 7th IEEE International Workshop on Storage Network Architectures and Parallel I/O (May 2011), 2011.

Raúl Rojas. Neural Networks: A Systematic Introduction. Springer-Verlag New York, Inc., New York, NY, USA, 1996.

Chris Ruemmler and John Wilkes. An introduction to disk drive modeling. IEEE Computer, 27:17–28, 1994.

Jan Fabian Schmid. Vorhersage von E/A-Leistung im Hochleistungsrechnen unter der Verwendung von neuronalen Netzen. Bachelor’s thesis, Universität Hamburg, 12 2015.

Lei Zhang, Guiquan Liu, Xuechen Zhang, Song Jiang, and Enhong Chen. Storage Device Performance Prediction with Selective Bagging Classification and Regression Tree. In Network and Parallel Computing, IFIP International Conference, NPC 2010, Zhengzhou, China, September 13-15, 2010. Proceedings, pages 121–133, 2010.




How to Cite

Schmid, J. F., & Kunkel, J. M. (2016). Predicting I/O Performance in HPC Using Artificial Neural Networks. Supercomputing Frontiers and Innovations, 3(3), 19–33.