The first result that one would want to know about a new protocol implementation would be the additional network overhead that it introduces. For this purpose, we compare the one-hop throughput of a Meghadoot network with that of a direct wireless transfer (without Meghadoot) between two nodes. In the case of a Meghadoot network, data transfer is performed between the IN and a node at a CZ radius of . TCP throughput is measured by transferring a 10MB file using ftp. The results are shown in Figure 5.1.
From Figure 5.1 one can infer that in a Mbps network the bottleneck is the network bandwidth, due to which the additional overhead involved in the Meghadoot protocol is not visible. As the network speed increases, this overhead becomes apparent. Moreover, with an increase in network speed, one can clearly observe the increase in the ratio between the througput with Meghadoot and the throughput without Meghadoot.
The motivation for implementing the Meghadoot architecture in kernel space was to enhance the throughput compared to a user space implementation. Hence, in this section we compare the performance of this kernel space implementation with that of the user space implementation.
The topology used was a linear one with three nodes. There is an IN with a CZ of radius two. TCP throughput is measured by transferring a MB file. Two measurements were made:
Figure 5.2 shows the throughput analysis for a Mbps network, while Figure 5.3 shows the throughput analysis for a Mbps network. It can be seen that an increase in the network bandwidth brings out the poor performance of the user space implementation, when compared to the kernel space implementation. Although the overhead in the user space implementation is high, it does not show up in a Mbps network as the network speed in the real bottleneck in such a network (low bandwidth), which is not the case in a Mbps network.
The rate of beaconing (AP Beacon, MN Beacon and Neighbour Update packets) can also have an impact on the throughput of a network, but this will come into play only in a network with a high node density. This could not be measured due to the lack of a sufficient number of nodes.
