All Products
Search
Document Center

Migrate the data of a world-leading media investment management company from Cloud Bigtable to Lindorm

Last Updated: Jul 14, 2021

Solution highlights: The maximum response latency for 99% of the requests is reduced to less than 5 milliseconds. The solution provides cross-zone disaster recovery capability and supports flexible data lifecycle management.

Customer feedback

We use ApsaraDB for Lindorm (Lindorm) because of two reasons. In terms of the product capability, Lindorm performs better than Google Cloud Bigtable. After the data is migrated to Lindorm, the request submission capability is more stable and the request glitch rate is much lower. In terms of the service quality, the Lindorm team has a very strong sense of service. Although our company is based in the United States, the Lindorm expert team always provides us supports in a timely manner. The Lindorm team has very strong technical capabilities and always solves problems in a short time.

Challenges

  • The total size of data is more than 10 TB, and the data size of a key-value pair can be megabytes.

  • A multi-level management system is required to manage the data lifecycle. The data lifecycle needs to be managed at different levels, such as table level, column-family level, row level, and cell level.

  • The business is sensitive to response latency. The maximum response latency for 99% of the requests must be less than 10 milliseconds.

  • The business requires strong consistency in multi-zone disaster recovery and has high service level agreement (SLA) requirements.

Solutions

  • The multi-zone disaster recovery capability of Lindorm ensures business continuity. When a data center fails, the disaster recovery capability ensures strong consistency and eventual consistency of data. Strong consistency in a disaster recovery solution means that the recovery time objective (RTO) is 60 seconds and the recovery point objective (RPO) is 0 seconds. Eventual consistency in a disaster recovery solution means that the RTO is less than 10 seconds and the RPO is less than 1 second.

High availability architecture for GroupM
  • Lindorm uses the CCSMap and BucketCache schemas developed by Alibaba Group and the Z Garbage Collector (ZGC) algorithm of Alibaba Dragonwell. This reduces the garbage collection (GC) pause to less than 5 milliseconds. The throughput is improved by 20% and the memory efficiency is improved by 40%. The response latency is reduced to less than 5% of the latency when the open source solution is used.

  • Lindorm uses a multi-level management system to manage the data lifecycle at different levels such as table level, column-family level, row level, and cell level. This helps the media investment company customize data lifecycle based on the business scenarios.

Benefits

  • After the data is migrated from Cloud Bigtable to Lindorm, the system handles business requests in a stable manner and the overall cost is also significantly reduced.

  • Lindorm provides multi-data center disaster recovery capability and an SLA that guarantees an uptime of 99.99% throughout the year.

  • Lindorm Tunnel Service (LTS) enables real-time data backup and allows you to export incremental data to heterogeneous databases.