Abstract: In this paper, we focused on the performance analysis of the buffer-aided D2D communication in cellular networks. Different from the previous studies of full traffic model, we profiled a tractable transmission model for D2D communication with dynamic data arrival. Considering the effect of the interference from cellular user and the channel fading, The D2D user data transmission process with a dynamic interference model was depicted by M/G/1 queuing model. Adaptive modulation and coding was used to divide signal-to-interference and noise ratio values regions which corresponded to a fixed data rate. In order to analyze the situation of data packets queuing up in the D2D user buffer, this paper formulated it as a Embedded Markov chain and computed its steady-state distribution. We adopt the adaptive transmission rate to make queuing analysis which need predict the SINR in certain intervals and use a dynamic traffic model. Finally, the s of performance metrics in terms of average queue length, mean throughput, average packet delay and packet dropping probability were obtained. Simulation results show the validity and rationality of the theoretical analysis.