Energy efficient computing offloading mechanism based on FPGA cluster for edge cloud

Towards the computing overload problem in cloud data center, we propose a computing offloading mechanism based on FPGA clusters for edge cloud. Firstly, the FPGA cluster we proposed is decoupled from traditional server, and FPGA BOX is used to power batch FPGA accelerators. The network-based FPGA cluster is deployed at the edge to undertake the computing tasks of the data center. On this basis, we propose an edge cloud network model based on FPGA clusters, and study the energy consumption minimization problem. Secondly, considering the diversity of computing tasks, the MapReduce algorithm based on the numbering mechanism proposed in this paper realizes the classification of computing tasks, and converts the energy consumption minimization problem into the energy consumption minimization problem when the user's computing tasks are determined. Based on the classification results, our improved Hungary algorithm can obtain the minimum energy consumption. Meantime, the computing task results can feedback to corresponding user according to the user information in classification results. Finally, numerical calculation results show that the computing offloading mechanism proposed in this paper shows good performance in terms of energy consumption.