This paper investigates the performance of the selective state space model Mamba in lifelong sequential recommendation. The goal is to model user behavior sequences over long periods, which is challenging due to computational complexity and the need to capture long-range dependencies. Mamba, a novel state space model with a selective mechanism, is evaluated for its effectiveness in this task. The authors introduce RecMamba, a framework that uses Mamba blocks to model user preferences over time. Experiments on two real-world datasets show that RecMamba outperforms existing models in terms of performance and efficiency. It achieves results comparable to the representative model SASRec while significantly reducing training time by 70% and memory costs by 80%. The results demonstrate that Mamba is effective in handling long user sequences, making it a promising solution for lifelong sequential recommendation. The study highlights the importance of considering longer sequences in recommendation systems, as they can capture richer interaction information and lead to more accurate recommendations. The findings also emphasize the efficiency of Mamba in processing long sequences, making it a viable option for building scalable and efficient recommendation systems. The paper concludes that Mamba is a promising approach for lifelong sequential recommendation, with potential for further research in areas such as multi-behavior recommendations and handling longer side information.This paper investigates the performance of the selective state space model Mamba in lifelong sequential recommendation. The goal is to model user behavior sequences over long periods, which is challenging due to computational complexity and the need to capture long-range dependencies. Mamba, a novel state space model with a selective mechanism, is evaluated for its effectiveness in this task. The authors introduce RecMamba, a framework that uses Mamba blocks to model user preferences over time. Experiments on two real-world datasets show that RecMamba outperforms existing models in terms of performance and efficiency. It achieves results comparable to the representative model SASRec while significantly reducing training time by 70% and memory costs by 80%. The results demonstrate that Mamba is effective in handling long user sequences, making it a promising solution for lifelong sequential recommendation. The study highlights the importance of considering longer sequences in recommendation systems, as they can capture richer interaction information and lead to more accurate recommendations. The findings also emphasize the efficiency of Mamba in processing long sequences, making it a viable option for building scalable and efficient recommendation systems. The paper concludes that Mamba is a promising approach for lifelong sequential recommendation, with potential for further research in areas such as multi-behavior recommendations and handling longer side information.