After purchasing the goods, use the account provided by the third-party platform to pay for the goods (to the third party), and the third party will notify the seller that the payment has been received and request delivery; the buyer receives the goods, inspects the goods, and confirms, Then notify the third party to pay; the third party then transfers the money to the seller’s account. Banks will not provide credit guarantees to merchants or buyers they are not familiar with based on their own risk considerations. This is an area that banks will not enter. Therefore, when the three-way payment license was just emerging, countless institutions were competing for it.

At the beginning of 2017, the People’s Bank of China issued new payment regulations clarifying the customer reserve funds generated by third-party payment institutions during the transaction process (that is, the pool of funds generated between the buyer’s payment and the seller’s collection). In the future, it will be deposited in a designated account, which will be supervised by the central bank, and payment institutions shall not embezzle or occupy the customer reserve funds. With the promulgation of this regulation, this has to a certain extent eliminated the channel for tripartite payment companies to embezzle the reserve funds for investment. The glorious tripartite payment has since cooled down and has become a subdivision that financial giants will only consider. market.

Sql or NoSql? This is a problem

After interpreting the logic and development history of the payment system itself, let’s talk about the core technology behind it. Considering the large transaction volume of Douyin live streaming and the need to support the spike scenario, a powerful database must be the core.

Currently, databases are basically divided into two types. One is a non-relational (NoSQL) database, which is a Key-Value database similar to Hadoop. This database is dedicated to massive data services, but requires data There can’t be too many correlation calculations between them, and there is almost no correlation calculation between each dynamic of related business users like Byte Video Social, so I believe that Byte should be relatively familiar to NoSql databases before; the other is in relational (Sql) database. For example, a customer’s transaction in an e-commerce scenario requires both the merchant’s inventory and the buyer’s account balance. In this scenario, a relational database is generally used for association calculation. And for spike scenarios like “Double Eleven”, for databases, it provides both massive data services and high-performance associated computing services. This obviously requires technology accumulation, and it opens up NoSql and Sql. In addition to Alibaba, it seems that only Tencent before the 2015 Red Packet War did it.

In the banking industry where the author works, transactions that involve changes in customer account balances are generally referred to as account transactions or financial transactions. The amount of WeChat red envelopes may not be large, but in fact, every red envelope grab by the user involves a financial Transaction, and such a large concurrent financial transaction, was completely unheard of in the industry before, and the key behind the red envelope is the support of the database. Before WeChat Pay, Tencent, like Byte, had basically no financial business scenarios. Compared with the previous social scenarios, financial transactions