Vertical E-commerce
A Drinking Water Manufacturer in China
About The Client
The client is a well-known consumer electronics, HVAC, robotics and automation systems, intelligent supply chain technology group, is the leading enterprise in China's home appliance industry. Its market effect, product quality and brand effect have always been among the best in China, and it is the forerunner of the emerging field of smart home.
Project Features
High concurrency requirement. Serving 1300 water stations including dozens of custom server staff, the system needs to support daily access from 20 million registered users and process over 3 million orders per day.
High stability requirement. Need to achieve 99.9% system availability during the year, and avoid no-response or downtime even in peak hours.
Containing a variety of buying scenarios. As per business requirements, sales promotion activities will be regularly carried out, such as panic buying and free event. System needs to handle users’ accesses with high concurrency by sales promotion to ensure the normal access during the promotion activities.
High network flow of images. When delivering the water or visiting distribution outlets, water delivery worker is required to take pictures as proof and send back to the server.
Order Polling Function. The App on water delivery worker’s phone will make an order polling every two minutes to update the list of current geographic locations to dynamically assign orders to the workers.
Service Value
Optimization of the system deployment architecture. Architecture separation of the business helped to reduce the stress of the network and concurrency, which improved the stability and responsiveness of the system.
Introduction of the image server, which reduced bandwidth consumption.
Introduction of cache mechanisms at all levels, which reduced unnecessary requests.
Optimization of the parameters of load balancing components and application services at all levels, enable the system to respond appropriately to the visits with high concurrency, and maximize hardware performance.
Various peak service scenarios were simulated to detect the system stability and responsiveness. Constantly optimize software and hardware performance to meet the system required performance.
On-line risks of the system were effectively avoided by digging out the potential memory leaks, deadlocks and other issues.
System performance was improved and all scenarios were ensured to meet or exceed design requirements. Most of the key scenarios can support more than 2000 concurrent users, compared with initially average 100 concurrent users.System throughput of certain scenarios was increased by more than 100 times.