This paper proposes a two-step deep reinforcement learning (DRL)-based service offloading (DSO) approach to optimize resource allocation and reduce computational costs in edge computing. The first step involves considering service and edge server costs during offloading, while the second step uses the R-retaliation method to evaluate resource factors and optimize resource sharing and subservice execution time (SET) fluctuations. The proposed DSO approach aims to minimize execution costs, transmission time, and server costs, while reducing deadline violation rates. The simulation results show that the DSO approach achieves lower execution costs and higher resource utilization compared to state-of-the-art methods, with reduced energy consumption. The paper also includes a detailed system model, problem formulation, and experimental results to validate the effectiveness of the proposed approach.This paper proposes a two-step deep reinforcement learning (DRL)-based service offloading (DSO) approach to optimize resource allocation and reduce computational costs in edge computing. The first step involves considering service and edge server costs during offloading, while the second step uses the R-retaliation method to evaluate resource factors and optimize resource sharing and subservice execution time (SET) fluctuations. The proposed DSO approach aims to minimize execution costs, transmission time, and server costs, while reducing deadline violation rates. The simulation results show that the DSO approach achieves lower execution costs and higher resource utilization compared to state-of-the-art methods, with reduced energy consumption. The paper also includes a detailed system model, problem formulation, and experimental results to validate the effectiveness of the proposed approach.