Accurate and fast time series anomaly detection is critical in various real-world applications such as industrial control. In recent years, diffusion models are employed for time series anomaly detection due to its effectiveness. However, traditional diffusion models are mostly of centralized architecture which limits their efficiency. Fortunately, with the rise of edge-cloud technology, it is feasible to design and deploy the diffusion models in a distributed manner which might significantly speed up the efficiency of diffusion models. To this end, a Distributed Diffusion-based framework for efficient time series Anomaly Detection (denoted as DisDiffAD for short) is proposed to facilitate the deployment and optimization of distributed diffusion models in edge-cloud environment. Specifically, a multi-objective optimization algorithm based on Langevin dynamics is designed to improve the efficiency of the distributed diffusion models in edge-cloud environment. The experimental results on 4 benchmark datasets show that DisDiffAD significantly outperforms all the baselines by reducing the comprehensive cost including inference delay, latency, and energy consumption by 5–15% while maintaining the detection accuracy degradation within 1%.