Using a Linux cluster to charge a billing system

Using a Linux cluster to charge a billing system

One of the most water-deficient areas in China is North China, and the most water-deficient area in North China is Shijiazhuang, “The Shijiazhuang City Water Conservation Office Leader’s Remarks Let the reporter feel the tension of Shijiazhuang water resources.

Thus, when Shijiazhuang Water Supply Company proposed to change the tradition of the past years, that is, the centralized collection of water user charging system, and to implement a refined water supply and billing model, it has been affirmed by all walks of life. However, technical problems have followed. For the past 5 to 60,000 large-scale large-scale charging models, the water supply company can use the simple method of “Windows+PC server+SQL Server”; once the meter reading, bank and network are realized The function of payment will be at least 600,000 households from the scale of users. Therefore, the overall environmental reform of the water supply company is imperative.

6000TpmC Demand

Shijiazhuang Water Supply Company's demand for new information systems is quite clear. Li Huawei, project leader of Shijiazhuang Water Supply Company, explained: “The new operating fee information system will meet the payment requirements of 600,000 account users every month. We hired a third-party consulting company to give us a brief calculation. According to industry experience, each window needs to serve 1000 users. At the peak of business, 50% of the business terminals simultaneously trade with the database host. Therefore, the number of business windows can be calculated. It will reach 600, and the number of concurrent windows at the peak is 300. Assuming that the response time of each transaction is no more than 3 seconds, the total number of concurrent transactions per minute will reach 6000, which is 6000TpmC.

From the perspective of network system structure, the new charging system is equivalent to one The metropolitan area network covers the information center computer room of the water supply company, the headquarters of the business office, five business branches, and bank collection points throughout the city. Moreover, since the usage rate of the toll collection system is mostly during the normal working hours during the day, there is no specific peak period within ten hours during the day, but the monthly end of the month and the beginning of the month belong to the peak period of payment, and the system usage rate is relatively high.

So, 6000 TpmC is the average value, and the data processing requirements during peak hours are higher. Therefore, the Shijiazhuang Water Supply Company proposed the demand of 15000 TpmC at the time of bidding, and the performance of the Unix server and the Dawning 4000A cluster participating in the bidding far exceeded this index. Therefore, when the performance does not become a bottleneck, the choice of the water supply company seems natural. Is to choose a more cost-effective platform.

Benefits of Linux Clusters

In the end, Shijiazhuang Water Supply Company chose 4-node Dawning 4000A as the hardware platform.

The water supply company believes that the core application of the billing system is database services. Most of the operations that cause large data throughput pressure on the database system are derived from statistics and query analysis (especially real-time query analysis). With the expansion of service scale, the system needs strong responsiveness, service processing capability, data security guarantee and high scalability. The Dawning 4000A server cluster is used as the main frame design to meet these requirements. The database server nodes are all equipped with four-way server Tiankuo R4280A, and equipped with Dawning DS3140F disk array products, combined with Redhat Advanced Server 3.0 operating system and Redhat Cluster Suit module, configured to become cluster mode, with oracle 9i rac to provide database services.

Li Huawei said that the total system software, server, storage and database combined investment is less than 1.7 million, compared with the performance of the Unix system is much lower.

As a system integrator, Shuguang believes that this system mainly solves two problems. First, the cluster architecture solves the problem that the server processing capacity is insufficient and the system work stability is not strong. In addition, the centralized storage system solves the problem. The problem of insufficient storage efficiency.