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.

70 Y. Ge, T. Kunz, and L. Lamontversions we study <strong>in</strong> this paper confirm this – QoS OLSR algorithms do enhance thenetwork QoS performance. However, <strong>in</strong> order to achieve this improvement, additional“protocol overhead” is also <strong>in</strong>troduced, which degrades the performance of these QoSrout<strong>in</strong>g protocols, especially with respect to “Packet Delivery Ratio” and “End-to-EndDelay” <strong>in</strong> high mobility cases. Does this then imply that we should abandon proactiveQoS rout<strong>in</strong>g and switch to on-demand QoS rout<strong>in</strong>g because of the cost? Not necessarily:Fig. 7. Average available bandwidth (<strong>in</strong> idle time) on the routes of the 4 OLSR algorithms– We do not know if on-demand rout<strong>in</strong>g algorithms have the same overhead problems.[3] discusses the performance of the “ticket-based prob<strong>in</strong>g” algorithm <strong>in</strong> a delay-constra<strong>in</strong>edenvironment, calculat<strong>in</strong>g what percentage of routes that the algorithmf<strong>in</strong>ds meet the delay request. But it fails to analyze other aspects of the rout<strong>in</strong>g algorithm,such as control overhead, packet delivery ratio etc. [11] tests the CEDAR algorithmus<strong>in</strong>g bandwidth as the QoS parameter, giv<strong>in</strong>g a detailed performance evaluation.However, [11] does not experiment with node movement. Nor does it run thesimulation <strong>in</strong> a real shared-channel environment, and the impact of channel <strong>in</strong>terferenceand packet collision are not considered.– Many proposed proactive QoS rout<strong>in</strong>g algorithm such as [10] and [7] just present abasic idea, without performance evaluation. So it is not clear whether the negativeeffect on the rout<strong>in</strong>g performance caused <strong>by</strong> the additional rout<strong>in</strong>g overhead is acommon problem to proactive QoS rout<strong>in</strong>g.Based on the above analysis, proactive QoS rout<strong>in</strong>g is still worth study<strong>in</strong>g. As theadded overhead is the ma<strong>in</strong> cost that affects the QoS rout<strong>in</strong>g algorithm’s performance,the future work on QoS rout<strong>in</strong>g <strong>in</strong> Ad-Hoc networks may be focused on how to reducethe overhead. Our future work plans <strong>in</strong>clude the follow<strong>in</strong>g:– TC packet collisions at the 2-hop neighbors cause the problem of stale rout<strong>in</strong>g tables.To avoid this problem, we can add some jitter mechanism <strong>in</strong>to the OLSR protocol– when an MPR receives a TC message, it waits for a random delay time before itrelays that TC message, <strong>in</strong>stead of relay<strong>in</strong>g it immediately.

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

Saved successfully!

Ooh no, something went wrong!