Introduce: “accept money to account 1 million” become the voice that ordinary businessman wants to hear most.
How long has it been since you touched cash?
The 47th Statistical Report on Internet Development in China released by China Internet Network Information Center (CNNIC) shows that as of December 2020, China’s mobile payment users reached 854 million. Data released by the People’s Bank of China shows that by the end of 2020, China had 123.22 billion mobile payment transactions, with a value of 432.16 trillion yuan.
These data are relevant to every one of us: breakfast on the way to work, parking in the neighborhood, a movie on holiday, a late-night snack at a snack stand… Mobile payment has already become an important infrastructure of society, and has formed the human fireworks in the streets and alleys. You must have heard the sound of “Take money and pay X yuan” issued by the machine.
Founded in 2013, it is the pioneer of the aggregated payment industry, which has penetrated millions of offline merchants and empowered hundreds of millions of people.
With the gradual maturity of mobile payment, a large amount of store transaction data has been deposited in the money bar. Make efficient use of these data to launch mobile payment and collection, marketing management software, advertising, supply chain, shared charging bank, merchant community and other services for merchants. Today, it has developed from an early aggregated payment service provider to a comprehensive service provider of digital stores, enabling millions of small and medium-sized businesses to realize the most basic digital transformation.
With hundreds of billions of data volumes, and more than 32 million transactions per day, how can you achieve massive storage and billing queries? Let’s go inside and collect the money and explore the story behind it.
The following is the fourth issue of “Super Visit” — How to help hundreds of millions of users keep their books in order, as Aliyun Database helps collect money? “.
1. Keeping the books in order for hundreds of millions of users
Zhang Yijia: How many merchants and the scale of users does it currently serve?
Chang Jianlong, CTO/ Senior Vice President of CTO: The service network of CTO has covered 660 cities in China, serving 4 million offline merchants, serving 32 million people per day on average and 20 billion people in total. From individual consumers to mom-and-pop stores to international chain brands, you can see the presence of CTO.
Zhang Yijia: How to store and manage the massive bills?
Director of Development: There are two key things that need to be captured. First, you need a data center to centrally store and manage your bill payments.
Secondly, it covers different types of customers, so we need to find ways to meet the differentiated needs of different users.
Therefore, Receive Money Bar has independently developed a hierarchical account system, which can make detailed statistics of the collection data of each brand, each branch and each cashier, clearly record each account, and customize the statement according to the financial requirements to solve the pain points concerned by customers.
Zhang Yijia: Payment is a very sensitive business, businesses and consumers are very concerned about their books. Collect money how to protect the user’s property safety, solve the user’s worries?
Mao Yu – Charge Bar Development Director: For property safety, Charge Bar emphasizes management and control. The boss can master the details without going to the store — remote account management, store management and personnel management. The APP provides daily bills, monthly bills and all collection records for easy reconciliation. APP can view all transaction information anytime and anywhere, which is convenient for cashiers and bosses.
Zhang Yijia: What were the biggest challenges in developing this system?
Yu Mao, Director of Development: There are three main challenges. The first is storage reliability, the second is storage cost, and the third is multi-dimensional composite real-time query with massive data.
Storage reliability is definitely important. Imagine if the bill is lost, the customer’s bill, the bank’s bill, the merchant’s bill and so on can not be checked and queried completely, this is unacceptable.
As for the storage cost, we have already stored hundreds of billions of data, and 32 million new data are added every day. This data increment is very large, and the storage cost is also very challenging.
In addition, different customers have different dimensions of data query needs. The query quantity of B side is relatively small, but the query dimension is complex. The query volume of C terminal users is relatively large, but the business logic is relatively simple.
For chain businesses, it is necessary to support multiple dimensions of brand, store, cashier, time and other random combination query; For consumers, you need to support billing queries based on a time range. In conclusion, it is necessary to support the random combined query of 30+ dimensions in order to meet the query needs of various types of users at present.
Zhang yijia: collect money how to solve these problems?
After a lot of research, analysis and testing, and comparing different technologies and products, we finally choose to use Aliyun’s cloud native multi-mode database Lindorm as the storage center to build the whole billing system. The system has been online and in operation for more than one year.
Lindorm was chosen for three main considerations: Lindorm is based on multi-copy technology, providing 99.99999999% data reliability, a total of 10 9s;
Lindorm has the maximum compression ratio of 10:1. Combined with the ability of cold and hot separation, the historical bill 6 months ago is converted into cold data for storage, which reduces the comprehensive storage cost by 2/3.
Lindorm provides an integrated bill retrieval solution, which can automatically create appropriate indexes against the bill data and support real-time queries with a random combination of more than 30 dimensions.
Zhang Yijia: Historical bills from 6 months ago are highly compressed and can be retrieved?
Director of Development: Yes. With the security of Aliyun Lindorm, mass storage and real-time query is no longer a contradiction.
2. Massive data to assist the decision of returning to work
Zhang Yijia: In 2020, because of the epidemic, offline stores are more or less facing business difficulties. Let’s collect money. As an enterprise serving offline merchants, what help does it provide to them?
Mao Yu – Development Director of Caoqianba: Caoqianba is committed to using the network and data power to help offline merchants operate their business. Based on the original traffic generated after payment, Caoqianba has launched the Tracking Data of National Offline Commercial Consumption Index, which can reflect the activity of offline commercial production in real time.
On the one hand, this data will be provided to the government to assist it in formulating policies to support the resumption of work and production of micro, small and medium-sized enterprises. On the other hand, open to our vast number of business users, to help businesses effectively develop a reasonable plan to return to work.
Zhang Yijia: How is the index calculated and how to ensure the accuracy and real-time performance of the data?
Mao Yu – Development Director of Caoqianba: Caoqianba’s business covers all walks of life of offline consumption. Based on the massive consumption data stored in Lindorm and certain econometric model, we can calculate and obtain the commercial consumption index in real time.
Zhang Yijia: Lindorm still plays an important role in this system?
Director of Development: Yes, the entire storage and computing of the system is built around Lindorm. All of our data are stored on Lindorm. Based on the integrated storage and computing solution provided by Lindorm, it can easily connect with the computing ecology, and help us quickly build a set of data processing architecture to support application services online.
Lindorm also provides wide table storage with KV structure, compatible with HDFS file storage, SQL query and data retrieval functions. Our source data and dimension tables are stored on Lindorm, and the calculated result table is also written back to Lindorm to provide the upper-level business query. Effectively avoid data migration and simplify system architecture.
With the Lindorm-built solution, the development rollout time was cut in half.
3. If you want to get rich in the digital age, make a good database
Zhang Yijia: In 2020, Aliyun released Lindorm. What are its characteristics?
Shen Chunhui, Senior Technical Expert of Aliyun Intelligence/Head of Lindorm R&D: Lindorm is the first independently developed NoSQL product released by Ali Cloud. Based on the cloud native architecture of storage and computing separation and multi-mode sharing and fusion, Lindorm supports low-cost storage and processing of massive data, supports rich data retrieval capabilities, supports one-button docking computing ecology, interactive, batch and stream processing and other computing processes. Help users quickly build a set of data processing architecture, in the Internet, the Internet of things in the scene in the immediate need.
Meanwhile, although Lindorm is independently developed, it firmly embraces open source ecology. The data access interface supports SQL, HDFS, HBase, CQL, OpenSDB and other open source standards. It is simple and easy to use without additional learning cost.
Zhang Yijia: The database independently developed by Ali is based on the demand of its own e-commerce business. From its own business to To B service, how does Ali Cloud maintain a keen insight into the demand?
Shen Chunhui, Senior Technical Expert of Aliyun Intelligence/Head of Lindorm Research and Development: From users to users, this is the invariable logic of product continuous iteration. In the process of To B service, in addition to the stability and cost demands, it is difficult for us to hear customers directly tell us what they want, which is a big challenge. And the database direct connection business, its demand is a large number of existence.
To this end, we mainly rely on two methods: one is the field visit. The database department has an expert service team, who will regularly visit customers for business and technical exchanges, timely understand the use of products and demands, and explore the efficiency optimization scheme of remote communication and collaboration;
The other is trend forecast, including industry trend, technology trend. For example, Lindorm has pre-invested a lot of technology research and development in the direction of the Internet of Things. The goal is to improve the processing efficiency of time series data and location data of terminal devices by orders of magnitude, which has a certain predictability in advance and will make timely adjustments with the development of the industry and application feedback.
This article is the original content of Aliyun, shall not be reproduced without permission.