Digital transformation practice of Zheshang Bank

The road to digital transformation of Zheshang Bank

Zheshang Bank is a relatively young company, , one of the twelve national joint-stock commercial banks . Zheshang Bank was established in 2004, so it is also a post-00 bank. In 2020, Zheshang Bank "The Banker" magazine Global Bank 1000 list , currently listed in Hong Kong and Shanghai.

Zheshang Bank started to comprehensively promote digital transformation very early. In 2010, Zheshang Bank launched electronic banking services and mobile banking services. In 2017, it launched its first blockchain service platform. In the same year, it released its direct banking brand. In 2018, Zheshang Bank's national standard A-level data center was opened, and in 2020, it will establish a financial technology subsidiary of Easy Enterprise Bank.

In the process of digital transformation, Zheshang Bank emphasizes three key points : The first is establish a digital thinking and concept . In the recently released "Fourth Five-Year Plan" of Zheshang Bank, it emphasized that digital capacity building is one of its competitiveness, and further established the leading role of financial technology. Innovative transformation of business model and management.

The second is the accumulation and application . In terms of data, Zheshang Bank has accumulated a lot of data. On the one hand, it promotes the integration of internal and external data, continuously accumulates and integrates internal and external data, and cooperates with external agencies such as the government and financial technology companies. introduces industry and commerce, justice, customs, etc. The external data actively connects with the Provincial Big Data Bureau and the Provincial Comprehensive Financial Service Platform . After several years of construction, Zheshang Bank has accumulated PB-level data and more than 10,000 data sheets . The second aspect is continue to deepen data applications , Zhejiang Bank established a cover the whole client, covering the entire process of wind warning and control decision-making model to achieve a risk detection, identification, processing, recording and sharing The comprehensive and three-dimensional monitoring of meets the differentiated and intelligent risk identification and control of 16139ccb326614 for public, retail, and

Zheshang Bank's big data control platform integrates 16139ccb326672 big data, knowledge , cloud computing 16139ccb326675 and other financial technologies to achieve customer access, relationship, post-loan management, early warning management, financial analysis, mobile applications, customer portraits And other functions, establish a full-process risk control and early warning system, and create a new credit risk management and control platform. The figure shows the main architecture of the big data sub-control platform. This platform has opened up loan credit approval and post-loan management , and realized early warning signal full life cycle management .

The third is strengthen technological innovation and application . The picture below is an innovative product of Zheshang Bank- receivable chain platform . Generally, companies have supply chains, and there are some core customers in the supply chain. Core customers have more accounts receivable and payable in the upstream and downstream settlements of the supply chain. On the one hand, a large number of accounts receivable occupy a serious amount of funds, resulting in greater liquidity pressure on core customers, and it is urgent to revitalize accounts receivable and reduce liabilities; on the other hand, core customers have more accounts payable to upstream suppliers. Affect the capital turnover of upstream small, medium and micro customers. Therefore, Zheshang Bank launched an innovative account receivable chain platform , using the technology 16139ccb3266f6, the decentralized and non- feature 16139ccb3266f7, to convert accounts receivable between customers into online circulation, repayment, Transfer and efficient and secure digital certificates. Customers into the platform can handle accounts receivable issuance, acceptance, confirmation, payments, transfers other business, effectively revitalize customer accounts receivable, reduce industry chain overall financial cost .

Based on the accounts receivable chain service platform, Zheshang Bank has formed a comprehensive financial service solution for different industries based on an innovative series of scenario application business models, combined with industry characteristics and customer area pain points, and has been applied in more than 20 industries.

16139ccb32677f in Zheshang Bank

Zheshang Bank built a open and easily expandable IT architecture . Infrastructure is basically cloud. also has private cloud, hosted cloud and proprietary cloud . The technical architecture layer is composed of enterprise-level SOA, self-developed distributed platforms, and the Internet of Things platform 16139ccb3267ba, which are used to support various platform-based businesses of Zheshang Bank, Traditional applications and Internet applications, etc., to achieve standardization, modularization and intelligence of products and services.

The first application scenario of is the 16139ccb3267ff anti-telecom fraud service . The feature of this system is that it needs to save all customer transaction data. There are currently several billion items. Banks need to feed back customer risks and transaction data to the public security organs. The traditional method is to run batches of generated result data through the big data platform and return it to the public security department. When a request comes, we need to fish out the data we want from the massive data. The problem with traditional big data platforms is that they cannot support high-concurrency index condition queries. A SQL query request is sent to the big data platform to run batches, and it takes tens of seconds to return. So in this case, this business cannot be done in real time. You should know that the degree of intrusion of synchronization or asynchronous to business logic is very different. Especially in this scenario, the core logic is a query , but developers need to do asynchronous, file sending, etc. in the middle. The data link is very long and the implementation of business logic is more complicated.

Therefore, we are looking for an ideal database that can provide fast queries. If a foreign stand-alone database platform is used for construction, the cost is relatively high. Zheshang Bank has several considerations in the selection of the database: the database cluster size and single table capacity unlimited , the performance is relatively excellent under large data volume, takes into account transaction and real-time analysis scenarios, and has financial peers Application cases, and also have an active open source ecology, rich tools and documents . Based on this, Zheshang Bank chose TiDB distributed database to implement the first business scenario at that time.

After going online, TiDB synchronizes approximately one million rows of data every day, such as anti-telecom fraud business queries, from the original return time of tens of seconds to milliseconds , which is very fast. We regard TiDB as an ordinary database, and queries can return results directly, and business logic development becomes very simple. In this scenario, we have also verified TiDB in three centers in two deployment and high availability capability of live .

The second scenario is the foreign exchange transaction management platform . Similar to the above scenario, the data volume of the system has reached billion , which is about 300 million. Previously, the Oracle database was used. This level of data needs to be implemented with a partition table. Those who do operation and maintenance must know that the maintenance of the partition table is very troublesome. After migrating to TiDB, I can intuitively feel that query and run batch performance has improved . There is no need to partition tables. Just build an index and check directly. It is very simple and a system can be completed. In this scenario, we verified the feasibility of channel business transactions and running batches on the TiDB platform.

TiDB is a very useful database product, and it also has column storage. We started to build distributed ODS scenario based on TiDB. With TiDB as the center, all the business data of various upstream OLTP systems are synchronized to TiDB through various data synchronization tools. There are many types of business databases, including DB2, Oracle, MySQL etc. TiDB can be connected to these databases, and its distributed architecture meets the continuous access of various upstream heterogeneous transaction databases without capacity restrictions. In this scenario, we applied TiCDC to connect the incremental changes of some tables to the Flink streaming computing platform through the change data capture framework. Based on the streaming platform, it can do some asynchronous real-time streaming calculations, which can be applied to the management cockpit system, and it can make real-time statistics on industry transactions and various business indicators of the whole bank. In addition, we to the mobile terminal to provide real-time report viewing for mobile users, and successively launched online points, public welfare games and other applications, gradually replacing the existing centralized ODS platform.

Next, let’s talk about the application of TiDB in the scenario. The blockchain is the leading product of Zheshang Bank. storage is that one node stores the data of the entire network, and the data of each node is the same, which is synchronized by the 16139ccb326980 distributed consensus algorithm Because the entire network is stored, as the amount of data increases, the test on storage becomes greater and greater. The database traditionally used by the blockchain system is LevelDB, which is a stand-alone Key-Value database with an upper limit of capacity. With the increase in transaction volume, we have tested that when the blockchain data reaches an order of magnitude, the performance of LevelDB drops significantly. Later we found that TiKV is also Key-Value database , which can match very well on the interface.

In this scenario, we did not use the complete TiDB, but used PD and TiKV to move the storage of the blockchain business to TiKV. platform has a storage interface designed to adapt to a variety of storage in the form of 16139ccb3269ab plug-in. TiDB’s professional service support staff gave us a lot of help during the adaptation process. After the storage is switched to TiKV, the effect is very obvious, effectively improving the throughput of the blockchain.

In general, TiKV helps the solve two problems : one is performance , and the other is the problem of capacity . When the transaction volume and storage volume reach a certain level, LevelDB will recover resources. First, the throughput of reading and writing will decrease, and secondly, when the recovery is performed, the performance curve will jitter. After TiKV went online, the test found that the effect was good, and the pressure was increased to a large magnitude. The performance of TiKV is relatively stable. According to the current test magnitude, there is no trend of performance degradation. According to the architecture design of TiKV, its carrying capacity is theoretically unlimited.

Next, share some thoughts on the construction of the Zheshang Bank's distributed data center, that is, the new data center. The following figure shows the current technical solution for the entire distributed data center. TiDB is applied to the distributed ODS scenario . The data of various transaction libraries of the whole bank is synchronized to TiDB. There are two roads behind, one is TiCDC and the other is TiSpark . TiCDC is connected to the distributed stream processing platform, TiSpark is connected to the distributed batch processing platform, and then connected to the big data platform. Using TiSpark to connect to Spark, the data is synchronized to HDFS big data processing, and the processing results of streams and batches are stored in A unified terminal storage (OLAP library). The data service layer of the data center is used to shield the storage difference. If it is near real-time data analysis, the trade will directly hit the traffic to TiFlash . If you need processed data , visit the OLAP library. .

depth cooperation to promote innovation in financial scenarios

Finally, review the course of cooperation between Zheshang Bank and TiDB. Starting from a distributed relational database, TiDB was tried and gradually expanded to data paste source storage and ODS monitoring. eliminates data islands and promotes data sharing . Zheshang Bank then applied TiDB to the bank's data center, using real-time data to drive business. We hope that TiDB will lay a solid foundation for data services, closely integrate business services and data to form a closed loop, and finally land in Zheshang Bank’s financial innovation scenarios, such as key transaction services, blockchain, etc.