13.07.2015 Views

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

58 M. Diha and S. PierreThe tunnel<strong>in</strong>g time is smaller <strong>in</strong> the proposed model compare to the Mobile IPmodel. The data process<strong>in</strong>g is faster because of the multiprocessor architecture. Themessages sent to the mobile users spend less time <strong>in</strong> the message queue on the HA. Inthe mobile IP model the message i + 1 will wait longer than the message i on themessage queue, <strong>in</strong>creas<strong>in</strong>g the process<strong>in</strong>g time. But <strong>in</strong> the proposed model the messagei + 1 can be process <strong>in</strong> parallel on a different processor while process<strong>in</strong>g themessage i on an other one. As a result the total time spent <strong>in</strong> the system is reduced.Overall the tunnel<strong>in</strong>g time is reduces <strong>by</strong> 10 % to 30% <strong>in</strong> the proposed model which isabove the targeted objective of 25%.4.3 Task Schedul<strong>in</strong>g and AssignmentFigure 10 shows the number of tasks miss<strong>in</strong>g their deadl<strong>in</strong>e for different number ofprocessors with different speeds follow<strong>in</strong>g Gaussian and exponential distributions.Missed deadl<strong>in</strong>es30201000 1000 2000 3000 4000Numebr of tasksFig. 10. Processors with different speeds distributionRate= meanRate = expo.V = gaussianThe number of tasks miss<strong>in</strong>g their deadl<strong>in</strong>e <strong>in</strong> the Gaussian distribution is lowercompare to the exponential distribution. The reason is that <strong>in</strong> the first case the speedsof the processors are close to the mean speed. It is the contrary <strong>in</strong> the exponential casewhere the distribution is larger with more low speeds. This leads to a higher ratioExecution Time/Processor Speed and number of missed deadl<strong>in</strong>es. So, for configurationswith different speeds, the speeds of the processors must follow a Gaussian distribution<strong>in</strong> order to have an optimal schedul<strong>in</strong>g and assignment for the tasks.5 ConclusionIn this paper we presented a Mobile IP architecture and mobility management algorithms<strong>in</strong> a real-time context. The implementation of the proposed architecture andalgorithms gave better results for the location update and tunnel<strong>in</strong>g average times aswell as the CMR compare to the exist<strong>in</strong>g architecture and algorithms. The locationupdate time is reduced <strong>by</strong> 20% to 80% while the tunnel<strong>in</strong>g time is reduced <strong>by</strong> 10% to30%. These results meet time constra<strong>in</strong>t <strong>in</strong> real-time systems. The multiprocessor

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!