Check out the new USENIX Web site. next up previous
Next: Heartbeat Period and Awareness Up: Simulation Results Previous: Effect of Leadership Handoff

Limits on Heartbeat Period

Next, we analyze the overhead of the algorithm by exploring the point at which it saturates the underlying network. This saturation point depends on the setting of the heartbeat period and the awareness horizon. It is fairly obvious that decreasing the heartbeat period results in more frequent communication between the nodes and therefore the ability to track faster targets. This is based simply on the speed of response necessary for faster targets. However, the bandwidth limitations in the wireless medium place limits on our timer settings and constrain our architecture's ability to track migrating events. To determine the bandwidth needs of our algorithm, we start with a very small leader heartbeat period (that saturates the network), then increase it gradually. We plot the resulting connection delay, which is the time it takes to send a message from the moving entity to the friendly-force entity. As the heartbeat period increases to the point when the network is no longer saturated with heartbeat traffic, we observe a sharp decrease in the connection delay. Figure 6 shows this effect. The experiment is repeated for different awareness horizons, expressed in the number of hops that leader heartbeats are propagated to. It is seen that when the horizon is increased, the onset of overload occurs earlier as more messages are communicated.

Figure 6: Effect of timer settings on message delay (speed = 12 m/s)
\begin{figure}\centerline{\psfig{figure=figures/D.eps,width=0.5\textwidth,angle=-90}}
\end{figure}

From Figure 6 we can see that the bandwidth of the wireless medium is fully saturated when the leader heartbeat period is reduced to approximately 2.9, 5.9, and 11.7 ms for an awareness horizon of = 1, 2, and 3 hops respectively. To conservatively avoid this saturation point and ensure enough bandwidth is left for alternate local traffic, we multiply these numbers by 5 (i.e., limit the worst case overhead of tracking to 20%). Hence, in the rest of the evaluation section, we consider only those leader heartbeat periods that are above 12.5, 25, and 50 ms for 1, 2, and 3 hops respectively. We next turn our attention to the selection of the leader heartbeat period and the awareness horizon, the two key parameters of the algorithm, subject to the above constraints.


next up previous
Next: Heartbeat Period and Awareness Up: Simulation Results Previous: Effect of Leadership Handoff
root 2003-03-05