Classifying Linux Memory Consumption Patterns Based on Self-Organizing Maps

• Mauricio T. N. G. Lin INdT
• Edjard de S. Mota UFAM
• Ilias Biris INdT

Abstract

The problem of memory exhaustion with restricted memory still sparks long discussions inside the Linux kernel developers community. The current implementation to treat the memory exhaustion in Linux is named as Out of Memory Killer and its algorithm to select processes for termination requires more investigation in terms of memory consumption behaviour. This paper describes a methodology to classify memory consumption patterns of Linux applications based on a neural network model known as Self-Organizing Maps. A tool was developed to provide the practical opportunity to apply this mechanism for the classification of memory consumption patterns related to real Linux applications use cases.

References

Borgelt, C. (2000). Self-organizing map training visualization. [link]. School of Computer Science - Otto-von-Guericke-University of Magdeburg.

Haykin, S. (1999). Neural Networks: A Comprehensive Foundation. Prentice Hall, 2nd edition.

Honkela, T. (1997). Self-Organizing Maps in Natural Language Processing. PhD thesis, Helsinki University of Technology - Neural Networks Research Centre, P.O. Box 2200 FIN-02015 HUT, FINLAND. [link].

Lin, M. (2006). Metodologia para classificação de padrões de consumo de memória no linux baseada em mapas auto-organizáveis. Dissertação de Mestrado, Programa de Pós-Graduação em Informática, Universidade Federal do Amazonas. [link].

Lin, M., Medeiros, V., Novellino, R., Biris, I., and Mota, E. (2005). Memory management approach for swapless embedded systems. Linux Journal, pages 36–43.

Singh, D. I. M. (2004). Embedded linux: The 2.6 kernel is ideal for specialized devices of all sizes. [link].