Reinforcement-Learning-Based Software-Defined Edge Task Allocation Algorithm

With the rapid growth in number of IoT devices at edge network, fast, flexible and secure computing has emerged, but disadvantage insufficient power servers is evident when dealing with massive tasks. To address this situation, firstly, a software-defined edge-computing architecture (SDEC) proposed, merging control layer computing, where multiple controllers share global information about network state through an east–west message exchange, providing for collaboration servers. Secondly, reinforcement-learning-based task allocation algorithm (RL-SDETA) proposed IoT, which enables to allocate computational tasks most appropriate execution based on they have obtained. Simulation results show that RL-SDETA can effectively reduce finding cost optimal server completion time its energy consumption compared various methods such as random uniform.