Enhancing the Efficiency of Memory Management in a Super-Paging Environment by AMSQM - Advanced Operating Systems and Kernel Applications - page 290

(such as: trace 1, trace 2, trace 5, trace 6, trace 7,
trace 8 and trace 11), AMSQM achieves a nota-
bly higher memory hit ratio, since contemporary
applications require a big portion of the memory
and reducing the number of the page faults in
such applications can significantly improve the
overall performance.

benchmarks that have a preference of other values
and there are also a small number of benchmarks
that will have a preference of adaptively modi-
fied values. Therefore, we believe that adaptively
modified values can improve the performance of
several benchmarks. Another issue that should
be addressed as well is the mutual influence of
the processes scheduled together (Wiseman and
Feitelson, 2003).
In addition, we would like to find a pattern
for super-pages reoccurrence. Such a pattern can
improve the efficiency of the super-page promo-
tion decisions. The traditional threshold parameter
seems to be insufficient for taking the most ben-
eficial decision. Some applications like (Wise-
man et al., 2004), (Wiseman, 2001), (Wiseman
and Klein, 2003) have a pattern of supper-pages
reoccurrence and the Operating System can take
an advantage of it.
The current results are encouraging and they
support our belief that the new page replacement
algorithm can notably enhance the memory man-
agement mechanism in the two above mentioned
manners: better TLB hit ratio and fewer page
faults.

concluSion and future worK

In conclusion, there is a need for a faster and
more suitable page replacement algorithm. AM-
SQM seems to meet this need. The speed and
super-paging suitability offered by AMSQM will
help page replacement algorithm avoid being a
performance bottleneck in computer systems for
years to come. AMSQM is an innovative adap-
tive page replacement algorithm for Super-paging
environment. It has been shown in this chapter that
AMSQM usually achieves a higher TLB coverage
than ARC and also a better page fault ratio in most
of the benchmarks that have been tested.
This chapter shows another important aspect of
the Super-Paging environment. We believe operat-
ing systems have had an improper attitude toward
the Super-Page replacement algorithm selection.
They usually just copy the old algorithms of the
traditional paging mechanism with no attention to
the new Super-Paging environment. This brings
about an improvement of the hardware support
for a smaller TLB miss ratio, but the software
support for a smaller TLB miss ratio is consider-
ably poorer.
So as to achieve an appropriate software sup-
port for Super-Paging environments, this chapter
has shown a way to adapt one of the most recent
algorithms to these Super-Paging environments
with the aim of obtaining a better TLB hit ratio.
In the future we would like to find methods to
dynamically set the AMSQM parameters (α,β,γ).
In the experiments that had been conducted in
this research, we have found that the values we
used for these parameters are the best for most
of benchmarks; however, there is a very few

acKnowledgment

The authors would like to thank SUN Microsys-
tems for their donation. Specifically, the authors
would like to express their sincere thankfulness
to Mr. Yosi Harel of SUN and Mr. Haim Zadok
whose help was above and beyond.

referenceS

Abouaissa, H., Delpeyroux, E., Wack, M., & De-
schizeaux, P. (1999). Modelling and integration
of resource communication in multimedia appli-
cations with high constraints using hierarchical
Petri nets. In Proceedings of IEEE International
Conference on Systems, Man, and Cybernetics
(SMC-99) , (Vol. 5, pp. 220-225), Tokyo, Japan.