The China Railway Construction Group’s big finance sharing platform based on Dameng’s database successfully provided a classic example of how Dameng’s database protection and read/write separation cluster replaced the use of the RAC cluster.

The China Railway Construction Corporation Limited (China Railway Construction for short) was established on November 5, 2007, and is a massive construction enterprise managed by the State-owned Assets Supervision and Administration Commission of the State Council. As a leading Chinese construction industry enterprise, it has contributed to government revenue and national development.

In recent years, China Railway Construction has repeatedly been selected in the Fortune Global 500 and the Top 225 International Contractors. It is one of the strongest and largest comprehensive construction groups in China and even the world.

China Railway Construction has over 20 engineering bureaus and each bureau has its own finance department and various independent financial systems. Their financial management was faced with continually high human resource costs and office expenses, overlapping system construction investment, difficulties in implementing unified management standards across each bureau, an inability to share financial data among different regions and other problems that hindered the group from carrying out efficient financial management. China Railway Construction’s single node database produced almost 500GB of archive logs per day, with almost 10GB of new data added daily. 60% of the transactions involved in realizing a centrally managed finance sharing platform are daily expenses, with high concurrency system requests mainly concentrated within and the remaining 40% of transactions primarily complex calculations, therefore this system needed to develop a customized database architecture for TP+AP services.

With the high concurrency and complex service logic of China Railway Construction's finance sharing service platform, the Dameng team decided to split the high concurrency expense control services and complex accounting services based on methodology after analysis and research.

With outstanding performance when used in the performance measurement process, DMDataWatch, based on DM7, helped realize the effects of read/write separation. The single-primary and multi-standby database, and slave machine was able to cope with 70% of read-only service diversion, helping to achieve a load balance between reimbursement and calculation services, and guaranteeing system reliability.

Expense control services were split according to demand levels and every five engineering bureaus’ expense control systems used a single master and slave, with read-only services diverted to the slave machine.

For reimbursement and calculation services with a large volume of complex SQL queries, Dameng added a slave machine to DMDataWatch. This slave machine supported the add, delete and change action to temporary tables which could divert complex logic reports to the slave machine. This solution was able to satisfy daily service and report service calculation demands and improve system performance.