DNN Application Oriented Migration Scheduling Strategy based on Genetic Algorithm

Migration Scheduling Strategy for Deep Neural Network (DNN) is a considerably popular method to migrate some complex neural network layers to the edge nodes having the rich resources for computation. In spite of this, this migration process causes additional time overhead and energy overhead and thus impacts the quality of experience for users. To this end, with the consideration of the DNN applications, this paper proposes a Genetic Algorithm (GA) based migration strategy to minimize response time and energy consumption for the multiple task scheduling condition. To be specific, the migration scheduling problem for DNN applications is formulated firstly. Then, the concrete strategy based on GA is devised. In addition, the simulation is implemented over the designated edge environment, and the experimental results demonstrate that the proposed strategy outperforms two baselines in terms of response time and energy consumption.