In this section, we highlight the performance of FU-Serve and discuss the results. The details of simulation parameters used are listed in
Table 1. We consider the presence of static and dynamic fog nodes in our work. We divide the simulation area into grids, and a static fog node is present at its center in each grid. For simulation, we consider four
. In the case of a dynamic fog node, a UAV is elected from the entire simulation area as mentioned in
Section V. The sensing range of each UAV is considered as 30m with an initial energy of
J. The location of the UAVs changes randomly within the simulation area. As some of the applications served by the FU-Serve are time-critical, we consider the presence of a partial amount of data generated from the applications to be critical. For our simulation, we consider
and
of data critical.
We consider the presence of static fog nodes in FU-Serve and compare them with traditional architecture in terms of transmission time.
Figure 2 presents the existence of
and
of critical data. In both cases, we observe that the transmission time is elevating with an increasing number of UAVs in the system. Additionally, we notice that the traditional UaaS architecture consumes more transmission time as compared to the FU-Serve. The possible reason for such a trend in the result is that, in the FU-Serve architecture, the fog nodes are present near the UAVs producing data. While in UaaS, all the data produced by the UAVs are transmitted to the cloud only, which consumes more time as compared to the UaaS.
Similar to
Figure 2,
Figure 3 depicts the performance of FU-Serve in the presence of dynamic fog nodes. In this experiment, the multiple fog UAVs move over different locations in the simulation area. Each of these fog UAVs is responsible for collecting data from its respective cluster as mentioned in
Section V. Therefore, in FU-Serve, the UAVs in a cluster transmit the critical data to a selected fog UAV within the cluster. In contrast, in traditional, the only option is to transmit critical and non-critical data to a centralized entity such as a cloud or a server. Consequently, we observe that FU-Serve outperforms in terms of transmission time compared to the traditional UaaS.
Figure 3 depicts the variations in transmission time in the presence of dynamic and static fog nodes while considering the critical data
and
. In both cases, we observe that the transmission time is also increasing with the increasing number of UAVs in the platform. Interestingly, we also notice that when the number of UAVs present in the system is 5, the transmission time is almost equal in both the cases of static and dynamic fog UAVs. Whereas when the number of UAVs increased to 10 and continued to 25, the transmission time in the presence of dynamic fog UAVs is less compared to static fog UAVs. The possible reason for such a trend in the plot is that – a dynamic fog node changes its location after a certain interval and forms a cluster, for which distance is an essential factor. Consequently, the UAVs which want to transmit the critical data obtain a fog UAV in its near vicinity in the cluster. On the other hand, the static fog node located at the center of the grid may be far away from the UAVs in the grid. Therefore, when a UAV wishes to transmit critical data, it takes more transmission time than in the presence of a dynamic fog UAV. Typically, UAVs are resource-constraint in nature. In FU-Serve, multiple UAVs collaboratively perform different operations, and a few serve as fog nodes. During every operation, the UAV consumes a significant amount of energy. Therefore, we consider the residual energy as an essential parameter in performance analysis.
Figure 4 shows the variations in average residual energy with the change in the number of UAVs in the network while considering
and
of critical data in the network. We notice that the residual energy in the case of
critical data is less in comparison to
of critical data in the network. However, we do not witness any specific increasing or decreasing pattern in the plot. From this plot, we infer that the energy consumption heavily depends on the position and volume of data, not on the number of UAVs present in the system. From
Figure 4, we infer that the energy required to process the critical data of
is higher than that required for processing the
critical data. The reason is that the energy consumption of a UAV depends on the amount of data it is processing. Therefore, for processing
of critical data, a UAV consumes more energy as compared to the processing of
critical data.
For selecting a UAV as a fog node, the fitness (
) denoted in Eq. (
12) is an important factor. One of the parameters to compute the fitness is the degree of connectedness (DC). Moreover,
.
Figure 5 witnessed the increasing amount of
with the increasing number of UAVs in the networks. In this experiment, we vary the values of DC between
and examine the change in
. We observe that with the increasing values of the total number of UAVs in the network and DCs,
is also increasing.