Abstract: ［Objective］ To advance the modernization of rail transit control systems and enhance their safety, reliability and operational efficiency, a cloud-based control system architecture for rail transit is designed. The aim is to effectively integrate and deploy both the safety-critical and non-safety-critical functions of rail transit system onto a cloud platform. ［Method］ A cross-regional distributed control is realized by connecting target controllers to cloud platform servers via optical fiber. Characteristics of the rail transit cloud control system are analyzed, safety requirements of the system are outlined across six aspects: time management, space management, shared resource management, operation mode management, degradation mode management and safety monitoring. ［Result & Conclusion］ The virtualization layers of the cloud platform must be designed to ensure spatial isolation, temporal isolation, fault isolation, as well as management, predictability, security, and confidentiality. The platform scheduling mechanism includes both time-triggered and event-driven approaches, each impacting the platform temporal constraints differently. The hierarchical scheduling of the cloud platform involves four resource-sharing protocols, which must be selected based on different tasks.