“Since migrating our fund system to the cloud, we have done testing and the performance results have exceeded our expectations. We can process 11,000 requests per second in real-time and clear 300 million transactions within 140 minutes.”
Alipay / Tianhong Asset Management Case Study
In June 2013, Tianhong Asset Management and Alipay jointly launched Yu'e Bao, a financial management and payment product. Within a few months of its launch, the Yu'e Bao fund surpassed RMB 100 billion with over 30 million users. It became the first fund in China's history to scale to that size.
As the Yu'e Bao fund was experiencing very fast growth, it required infrastructure that not only scaled but that was capable of fast deployment without any downtime.
However, given that Yu’e Bao operates in the financial services industry where information security is vital, Yu'e Bao could not afford to compromise its security with the implementation of new infrastructure. Yu'e Bao therefore required a highly scalable model that could keep pace with volume increments and also provide more advanced security standards.
Finally, as a new fund Yu'e Bao could not afford to make huge upfront investments on infrastructure.
Why Alibaba Cloud
Alibaba Cloud was the logical choice to provide the infrastructure and migration that Yu'e Bao needed. Though Alibaba Cloud is affiliated to Alipay under the umbrella of the Alibaba Group, an affiliation alone does not guarantee a smooth migration.
Alibaba Cloud's efforts in then working with Yu'e Bao to systematically meet its infrastructure needs proved to be the right choice. Alibaba Cloud fulfilled the requirements and deployment was completed with just a few clicks. The infrastructure was up and running within a few days as opposed to months as would be the case with traditional infrastructure.
Alibaba Cloud's automatic partitioning through ApsaraDB for RDS was integral in underwriting data security, as well as the high level of built-in security in other Alibaba Cloud services, including identity and access management at resource and network level.
Lastly, the Alibaba Cloud Pay-As-You-Go model saved Yu'e Bao from needing to make a huge upfront investment in provisioning powerful infrastructure to meet its business demands. At the same time, the built-in elasticity enables Yu'e Bao to turn off idle resources and save costs.
Yu'e Bao went on to become the first successful fund to integrate cloud computing to manage direct sales and clearance systems. The innovative business model and technology architecture of Yu'e Bao has had a great influence on the finance industry and has become a benchmark in Internet financing.
After deploying to the cloud, the direct sales system was supporting over 1,000 subscribers per second and the clearance system was completing 300 million transactions per 150 minutes. The existing architecture though could not support the growth and volume. Yu'e Bao therefore re-designed the architecture to support massive concurrency to achieve system stability, reliability and build greater fault tolerance.
The redesigned architecture was therefore built on massive server clusters. The application layer is built on a cluster of servers followed by a MySQL database cluster at the backend to achieve the desired level of performance. Alibaba Cloud Server Load Balancer manages and directs traffic, and CloudMonitor performs fault detection.
“Based on the Alibaba Cloud platform, we successfully constructed the required cloud computing environment within a few hours. Since moving to the cloud, Yu'e Bao has delivered great clearance results. Clearance work that once took 8 hours can now be completed in just 30 minutes. If we had constructed such a large production environment on our own, we would have been forced to invest tens of millions of RMB. However, with Alibaba Cloud, this was all accomplished with just a very small investment.
Since migrating our fund system to the cloud, we have conducted testing and the test performance results have exceeded our expectations. We can process 11,000 requests per second in real-time and clear 300 million transactions within 140 minutes. In less than two months, the system also passed the peak traffic test with a stable and reliable performance.”