QLEN: A Load Distribution Algorithm to Improve QoS Factors Among Fog Devices

This article entails the proposition of a heuristic multi-objective algorithm for load distribution among fog devices. The decision for routing is made based on a weighted heuristic. The proposed algorithm, QLEN aims to improve three quality of service factors, namely L (load balancing efficiency), E (energy consumption) and N (network usage); hence the name QLEN. The importance of these parameters can be varied according to the application. This is done by adjusting the weights to suit the needs of the incoming tasks. For the default use case, the weights are tuned for the optimization of all three of the parameters. These optimal weights are calculated using stochastic gradient descent. QLEN is compared against fundamental load balancing algorithms such as Round Robin, Throttled Round Robin and Priority Based Load Balancing algorithm. The environment simulates a health monitoring application, using the iFogSim simulator. The results show that QLEN consistently outperforms the compared algorithms with respect to the targeted parameters. The efficiency of load distribution achieved by QLEN on average performs 1.69% better than all the three algorithms; at the same time showing 54.86% lesser network usage, 3413.33 MJ lesser energy consumption and 10.39% reduced redirection to the cloud. The algorithm also scales efficiently, showing better performance as the number of fog devices increases.