Next, we explore the effect of the leadership handoff mechanism on performance. While our architecture includes a leadership handoff mechanism that explicitly notifies entity members when a new leader should be elected, we compare that mechanism to the case where the old leader simply dies in which case a timeout must elapse before the new leader election starts. Figures 4 and 5 compare the average number of entities formed around the target for different target speeds with and without the handoff mechanism respectively. For each target speed, we vary the leader heartbeat period. Larger periods mean that more time will elapse before leader failure is noticed. Observe that the curves in Figure 4 are generally lower than in Figure 5, indicating a larger fraction of successful runs. Remember that each point on each curve is the average of 10 runs. Points indicating a single formed entity mean that all 10 runs were successful. The handoff mechanism is more successful in tracking since it avoids the extra delay in leadership handoff making it less likely that the target will move to where it can be sensed outside of the awareness horizon, thus causing a spurious entity to emerge. Thus, in the rest of our analysis we only consider simulations where our handoff mechanism is present.
In addition, it is interesting how the choice of the heartbeat period strongly influences our architecture's ability to track an event. Slow periods will result in a slower transition resulting in the event migrating beyond the awareness horizon. Fast periods result in a congested channel which increases message loss and prevents nodes from hearing about an approaching event. In between these extremes, an optimal choice of the heartbeat period can be made. This choice will be investigated in more detail later in the section.
