Performance Bottleneck Analysis of Drone Computation Offloading to a Shared Fog Node

Computing in drones has recently become popular for various real-world applications. To assure the performance and reliability of drone computing, systems can also adopt computation offloading to a nearby fog or edge server through a wireless network. As the offloading performance is significantly affected by the amount of workload, the network stability, and the competing use of a shared resource, performance estimation is essential for such systems. In this paper, we analyze the performance bottleneck of a drone system consisting of multiple drones that offload the tasks to a shared fog node. We investigate how resource conflict due to computation offloading causes the performance bottleneck of the drone computation system. To model the behavior of the system and analyze the performance and availability, we use Stochastic Reward Nets (SRNs). Through the numerical experiments, we confirm that the benefit of computation offloading deteriorates as the number of competing drones increases. To overcome the performance bottleneck, we also discuss potential solutions to mitigate the issue of a shared fog node.