Solution highlights: No Oracle databases are used. The architecture integrates the ApsaraDB for Lindorm (Lindorm) wide table engine, Lindorm Tunnel Service (LTS), and the Lindorm search engine.
Workloads such as managing logistics order and handling data from barcode scanners do not require transactions. We originally used Oracle databases to store such business data due to the concern of performance and stability. Then, we found that Cainiao Logistics uses Lindorm to handle such data. After we tested the architecture that integrates the Lindorm wide table engine and the Lindorm search engine, we found that Lindorm provides linear scalability. This makes Lindorm more suitable for logistics business. We do not need to modify the code or migrate data to new instances to handle data during rapid business growth and promotions such as Double 11. Lindorm databases are also more cost-effective than Oracle databases. We have used Lindorm for nearly two years. During this period, Lindorm has proved its stability and high availability. We are grateful for this service.
STO Express was founded in 1993. It is committed to building and developing Chinese brands. STO Express constantly improves the system that integrates the terminal network, the transit transportation network, and the information network. The company is rooted in the traditional courier business and has engaged in the field of e-commerce. STO Express provides professional services and strict quality management to promote the development of the courier industry in China. After more than 20 years of development, STO Express has constructed a comprehensive and smooth self-operated courier network in China. Up to January 2021, STO Express runs more than 4,500 branches and independent outlets and has more than 25,000 logistics centers. More than 300,000 people are working for STO Express.
A large amount of data is generated every day by workloads such as scanning waybills by barcode scanners, placing orders, and distributing orders. The amount of data generated by barcode scanners reaches hundreds of millions per day. The amount of data generated by placing and distributing orders reaches tens of millions per day. Therefore, the total size of data per day is more than 100 TB.
Different queries are required in different scenarios, such as exact queries based on order numbers or other dimensions, and fuzzy queries.
The amount of business data surges during business peaks such as Double 11 every year.
Lindorm is developed and maintained by a top technical team in China and provides stable and reliable services. It has been widely applied to multiple key services of Alibaba Group. Its stability and reliability have been well proven.
Lindorm is a cloud native multi-model database service that supports a wide table engine and a search engine. Lindorm provides Lindorm Tunnel Service (LTS) that can synchronize data in a wide table database to a database that is connected to the Lindorm search engine in real time. This ensures the consistency between the data stored in the wide table database and the search engine database.
Lindorm databases can be vertically and horizontally scaled to meet the elastic resource requirements during business peaks such as Double 11.
Lindorm uses an architecture in which storage is decoupled from computing. Storage nodes and compute nodes can be scaled based on business requirements and scaling operations do not interrupt the services of applications. Data and requests are automatically balanced. Lindorm frees STO Express from O&M. STO Express does not need to modify the code of its applications to use Lindorm.
The following figure shows the system architecture.
Lindorm helped STO Express migrate all the data from Oracle databases to Alibaba Cloud at a low cost. The multiple types of storage media, high compression ratio, and linear scalability provided by Lindorm ensure the scalability and high performance of the system.
The Lindorm wide table engine works with the Lindorm search engine to meet the data query requirements in different scenarios, such as fast searches for orders, waybills, and sub-orders, as well as multi-dimensional searches and fuzzy searches. The wide table engine supports 100,000 requests per second. Lindorm provides a query performance level five times higher than Oracle. The Lindorm search engine allows STO Express to run random multi-dimensional queries from tens of thousands of service outlets, and export and download the result sets in real time.
LTS efficiently synchronizes data from the wide table database to the search engine database in real time. This way, the business system does not need to write data to the two databases to ensure data consistency.
Lindorm uses a cloud native distributed architecture in which computing is decoupled from storage. Therefore, the Lindorm resources of STO Express can be elastically scaled on demand. This helps STO Express handle data in business peaks such as Double 11.