Major e-commence promotion events in China and other regions, such as the Double 11 and 618 events, have become a routine in the e-commerce industry. Market players are confronted with serious challenges of designing proper application and database architectures to handle traffic surges and provide customers with a smooth shopping experience during peak hours. With its ability to scale online and process highly concurrent requests with low latency, OceanBase Database has powered the Double 11 event for 10 years and allowed e-commerce companies to quickly and transparently handle business load fluctuations and improve system throughput during flash sales.
As e-commerce platforms launch promotions all year round, accurately predicting business growth and transaction traffic fluctuations in marketing activities is challenging. Therefore, e-commerce companies require databases that are linearly and smoothly scalable.
Given the poor scalability of conventional databases, project teams often start preparing for capacity scaling and rehearse plans long before a major promotion event goes online. This requires a huge investment in resources and manpower.
Conventional database scaling architecture often involves database and table sharding, which requires an enterprise to buy a large number of servers for different instances in advance. Moreover, a scaled-up database is difficult to scale down.
Stability is the top priority. Any jitter during peak hours can have a significant impact on the business of an enterprise and even result in asset loss. In addition, sharding and scaling-up operations may affect database stability.
Scenarios such as inventory deduction, flash sales, and bulk transactions can pose serious challenges in processing highly concurrent requests in hotspot rows of a database, which directly affects the system throughput and business revenue.
When a promotion event starts, users can increase the maximum values of resource specifications for a specific business or horizontally scale out the cluster by adding replicas with a few clicks on OceanBase Cloud Platform (OCP). The scaling is transparent to business applications, and the database throughput linearly increases with the resources, providing theoretically infinite processing capacity. When the promotion event ends, users can quickly scale in the database and release resources. OceanBase Cloud supports scaling of tenants, nodes, and clusters to allow users to develop flexible scaling policies as needed and cope with traffic fluctuations at low costs.
Upgrade of tenant specifications: Users can adjust the computing resources of a tenant without changing the total resources of the cluster. The upgrade takes effect immediately after the modification and is transparent to the business.
Vertical scaling of nodes: Users can increase the amount of available resources by scaling up the hardware specifications of a single node while the business is online. OceanBase Cloud supports the scaling of multiple tenants at the same time.
Horizontal scaling of clusters: When the resources of a tenant are scaled to the maximum hardware specifications of a single node, users can increase the number of resource units of the tenant by adding nodes to the cluster. As the data and computing resources of the tenant are distributed across multiple nodes, the cluster can achieve theoretically infinite horizontal scalability.
OceanBase Database has withstood extreme business scenarios of Alipay and MYbank during the Double 11 event in the past 10 years and delivered services with guaranteed performance and stability.
OceanBase Database supports fast replica switchovers and automatic load balancing, allowing users to flexibly use cloud resources and quickly scale out databases. After the peak hours, users can scale in (shrink) the databases without business interruptions. This greatly reduces the costs for major promotions.
Native partition scalability delivers linear throughput growth as resources increase. OCP enables online resource scaling for specific services with a few clicks. Replica switchovers happen automatically and transparently, allowing quick response to load fluctuations.