Intelligent Deep Reinforcement Learning Based Resource Allocation in Fog Network

Abstract

The increase in the smart devices have brought about the intelligence to the applications at the edge. The current day devices are rich in computations and communications which are in most cases are left under-utilized. These devices can be used to provide edge intelligence when deployed for computations on the streaming data from the IoT devices. To reduce the latency of sending the data to the cloud in case of real-time applications, a new paradigm of edge computing was introduced which meets the emerging challenge of handling latency aware applications on the fly. The other research issue to be handled with utmost importance is the load balancing which can be realized effectively by proper proactive resource allocation. Our proposed work involves the construction of SDN based Fog infrastructure wherein the research issue was tested and evaluated. The proposed Deep Reinforcement Learning algorithm helps in intelligent action selection based on the past experience data and the dynamic network parameters. Finally, the proposed work was compared with the state of the art OSPF algorithm in terms of service time and the load variance on increasing task allocation