Abstract: With the development of railway electronic payment platform business and the substantial increase in ticket sales and payment for the 12306 Internet ticketing and reservation system, the payment system has encountered bottlenecks in the transaction log and snapshot link. This paper studied the use of spark, Kafka, HBase and other key technologies, and designed data processing architecture based on Hadoop platform and Java development tools to meet the requirements of high performance and processing based on double active centers processing. Through practical application, it greatly improves the system processing capacity and better supports the development needs of railway business system.