Aiming at the problems of unreasonable micro-service splitting, poor reliability and bottleneck of large-scale data processing in the traditional scheduling cloud data integration architecture, a general distributed architecture solution based on the scheduling cloud platform is proposed. Firstly, the typical architecture of cloud platform is constructed, and the cloud platform components are effectively integrated with the actual business system based on the cloud computing concept.Secondly, the key distributed technologies are explored and studied, including metadata based file distributed storage, multi-executor distributed task processing center, data agent based distributed database storage, business module distributed deployment and other architectures. A comprehensive load scheduling algorithm based on exception task recovery mechanism is proposed. Finally, the efficiency of massive data processing, processing and access, and the performance of the system under high concurrency scenarios are improved under the proposed distributed architecture through real scenarios verification.