Ad Hoc Networks : Technologies and Protocols - University of ...
Ad Hoc Networks : Technologies and Protocols - University of ...
Ad Hoc Networks : Technologies and Protocols - University of ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
144 Transport Layer <strong>Protocols</strong> in <strong>Ad</strong> <strong>Hoc</strong> <strong>Networks</strong><br />
The proactive route error mechanism can prove to be disadvantageous<br />
when a link failure has occurred due to congestion. Consider an example<br />
where a link between nodes A <strong>and</strong> B is traversed by 2 TCP connections<br />
<strong>and</strong> In the default set-up, when a packet belonging to experiences<br />
congestion related link failure, only would be informed <strong>of</strong> the<br />
link failure prompting to choose a different route <strong>and</strong> thus relieving<br />
congestion along the original path for However, when the proactive<br />
route error mechanism is used, both <strong>and</strong> will be informed <strong>of</strong> the<br />
route failure making both <strong>of</strong> them to recompute their routes (although the<br />
same path might be chosen all over again). However, the characteristic<br />
<strong>of</strong> the default set-up to let route requests through in preference to data<br />
packets results in routes being chosen irrespective <strong>of</strong> the congestion along<br />
the path. Hence, in the example considered there is nothing to prevent<br />
flow from choosing the same path again even under the default set-up.<br />
Performance<br />
While the symmetric route pinning mechanism reduces the probability <strong>of</strong><br />
route failures, the route failure prediction mechanism reduces the occurrence<br />
<strong>of</strong> route failures by predicting them proactively. These two mechanisms directly<br />
reduce the number <strong>of</strong> retranmission timeouts in TCP. However, when a route<br />
failure does occur, the proactive route error mechanism reduces the delay in<br />
informing the sources <strong>of</strong> the route failure, thereby reducing the probability <strong>of</strong> a<br />
retransmission timeout. Hence, these three mechanisms in concert reduce the<br />
number <strong>of</strong> retransmission timeouts experienced by the connection as observed<br />
in Figure 5.7(a) when compared to the default TCP case in Figure 5.2(c) for<br />
the 1 connection scenario. A direct benefit <strong>of</strong> the reduction in the number<br />
<strong>of</strong> timeouts, is the resulting reduction in the loss percentage <strong>of</strong> packets <strong>and</strong><br />
the consequent increase in throughput as observed in Figures 5.7 (b) <strong>and</strong> (c)<br />
respectively.<br />
Trade-<strong>of</strong>fs<br />
The lower layer mechanisms in the ATRA framework help improve TCP’s<br />
performance without requiring any changes to the transport layer, by appropriately<br />
taking actions to mask out the negative impacts <strong>of</strong> route failures caused due<br />
to mobility. However, this is achieved at the cost <strong>of</strong> lower layers being required<br />
to be TCP-aware in their operations. Also, the route prediction mechanism<br />
relies on the signal strength <strong>of</strong> the received packets, it is challenge to make<br />
such predictions accurate under conditions <strong>of</strong> signal fading due to obstacles,<br />
multipath, etc. Finally, several characteristics <strong>of</strong> TCP that are by themselves<br />
inappropriate for operation over ad-hoc networks, are not addressed by this<br />
framework.