Adaptive Thresholds for Task Offloading in IoT-Edge-Cloud Networks

Abstract

The increased number of connected end devices requires carrying the generated loads efficiently. Cloud and edge computing are considered key technologies to overcome the limitation of end devices from many perspectives: computational power, storage capacity, as well as energy consumption. End devices can offload their overflow tasks to the cloud or to the edge for processing, storage, or analysis. This paper proposes an offloading queueing model and an adaptive offloading algorithm to enhance the offloading performance by making efficient offloading decisions. The suggested queueing model considered general service time distributions and offloading decisions are based on a processing time threshold values. The end device is modeled as an M/G/1 queueing system, while the edge node and the cloud are both modeled as M/G/m queueing systems. The proposed adaptive offloading algorithm can dynamically adjust and determine the perfect offloading threshold value that maintains a constant level of load on the resource limited end devices and edge node along with the increased/decreased loads. Simulation results prove the effectiveness of the proposed adaptive offloading algorithm in minimizing the task mean response times and balancing the generated loads efficiently.