All Products
Search
Document Center

Migrate the real-time bidding data of an interactive advertising company to Lindorm

Last Updated: Jul 14, 2021

Solution highlights: Data is migrated from a self-managed HBase cluster deployed on a data center to ApsaraDB for HBase Standard Edition, and then to ApsaraDB for Lindorm (Lindorm). This way, the resource overhead is reduced, and the maximum response latency for 99.9% of the requests is reduced by 90%.

Challenges

  • Real-time bidding (RTB) is different from the traditional advertising mode. In RTB mode, the submission latency of bidding requests for advertising exposure must be less than 100 milliseconds. If a bidding request is not submitted within 100 milliseconds, the advertising opportunity is lost and the company loses money.

  • To implement precise advertising, the RTB system has high concurrency requirements for accessing a large amount of user persona data.

  • The RTB system needs to collect and store user behavior data in real time. The user behavior data is generated when users respond to different advertisements. After the user behavior data is collected, the RTB system collects statistics based on different dimensions and different metrics. This way, advertising effectiveness can be obtained in real time and advertising strategies can be optimized based on the statistics.

Solutions

  • The company migrated data from a self-managed HBase cluster deployed on a data center to ApsaraDB for HBase Standard Edition. This reduces the resource overhead and the maximum response time for 99.9% of the requests. This way, the requirements of the advertising company are met.

  • The company then migrated data from ApsaraDB for HBase Standard Edition instances to Lindorm instances of the same specifications. This reduces the maximum response latency for 99.9% of the requests by 90%.

Storage and response latency optimization by using Lindorm Log Consensus (LLC) Garbage collection (GC) optimization provided by Lindorm

  • Lindorm uses a dual service architecture to provide a high availability solution. In this architecture, the primary and secondary clusters are deployed in the same region.

HBase Dual Service

Benefits

  • After the data is migrated from the self-managed HBase cluster deployed on a data center to ApsaraDB for HBase Standard Edition, the resource overhead is reduced. Alibaba Cloud provides service level agreement (SLA) guarantees and a complete portfolio of data management services to free the advertising company from O&M. Alibaba Cloud also offers a professional Lindorm expert team to provide free technical support. This helps the advertising company focus on their business development.

  • After the data is migrated to Lindorm, the response time (RT) for business requests is more stable. The maximum response latency for 99.9% of the requests is reduced by 90%. This effectively helps the advertising company increase their revenue.

  • The primary and secondary clusters are deployed in the same region for automatic failover. When the primary cluster for the advertising business fails, the workloads on the failed primary node are immediately switched to a secondary node.

  • The dual service technology helps the company effectively use the resources of the secondary cluster. This improves the stability of the bidding request submission system.