计算机技术与发展2025，Vol.35Issue(11)：20-27,8.DOI:10.20165/j.cnki.ISSN1673-629X.2025.0165
一种基于联邦架构融合扩散模型的云迁移方法
A Cloud Migration Method Based on Federated Architecture Integrating Diffusion Models
摘要
Abstract
Deep Reinforcement Learning(DRL)has shown great potential in the effective implementation of cloud migration.However,existing DRL-based cloud migration methods face issues such as high migration costs and instability.To address these challenges,we propose a Deep Reinforcement Learning method based on federated architecture integrating diffusion models(Fed-DMDRL).Fed-DMDRL is built on a federated architecture consisting of a central server and multiple clients,where each client uses a diffusion model-based deep reinforcement learning method(DMDRL)to train data locally.The diffusion model employs a transformer to generate more compact and efficient data representations,thus reducing unnecessary additional costs.The trained results are then aggregated on the central server,enabling stable data transmission through the federated architecture's distributed operation mode and not directly accessing raw data.Experimental results demonstrate that Fed-DMDRL significantly reduces the cost,interruption frequency,and response delay compared to the state-of-art methods,contributing to the effective implementation of cloud migration.关键词
深度强化学习/扩散模型/联邦学习/云迁移/分布式运行Key words
deep reinforcement learning/diffusion model/federated learning/cloud migration/distributed operation分类
信息技术与安全科学引用本文复制引用
檀龙伟,王丽芳,秦品乐,李沛聪,贾晗铭,黄昱帆..一种基于联邦架构融合扩散模型的云迁移方法[J].计算机技术与发展,2025,35(11):20-27,8.基金项目
山西省科技创新计划项目(20210222) （20210222）
山西省重点研发计划项目(202202010101008,202102010101011) （202202010101008,202102010101011）
山西省省筹资金资助回国留学人员科研项目(2024-118) （2024-118）
