Evaluation of market operating systems behavior under thrashing conditions

• Douglas Santos PUCPR
• Carlos Maziero PUCPR

Abstract

Virtual memory allows offering to processes more memory than that physically available in the system, using a disc as a memory extension. When there is not enough RAM memory to satisfy the processes, the system becomes slow, because it spends the time doing paging, characterizing memory thrashing. This paper presents an evaluation of some commodity operating systems under thrashing. A portable benchmark tool was developed to bring each system to a thrashing and then back to normality. We also identify the performance data available in each system and the mechanisms available to collect them.

References

bandwidth (2008). Bandwidth. [link].

Cowardin, J. (1997). A proc buffer for kernel instrumentation. Master’s thesis, The College of William & Mary.

Denning, P. J. (1968). “Thrashing: its causes and prevention”. In AFIPS Fall Joint Computer Conference.

Henning, J. L. (2006). SPEC CPU2006 Benchmark Descriptions. [link].

Jiang, S. and Zhang, X. (2005). “Token-ordered LRU: an effective page replacement policy and its implementation in Linux systems”. volume 60, pages 5 – 29.

Mauro, J. and McDougall, R. (2001). Solaris Internals: Core Kernel Components. Prentice Hall PTR.

McKusick, M. K. and Neville-Neil, G. V. (2004). The Design and Implementation of the FreeBSD Operating System. Pearson Education.

McVoy, L. W. and Staelin, C. (1996). “LMbench: Portable tools for performance analysis”. In USENIX Annual Technical Conference, pages 279–294.

Mucci, P. J., London, K., and Thurman, J. (1998). The CacheBench Report. [link].

nbench (2008). nbench. [link].

Russinovich, M. and Solomon, D. (2004). Microsoft Windows Internals, Fourth Edition. Microsoft Press.

Stream (2008). STREAM bench. [link].