Cut total costs by 86% and extend the data retention period by 4 times — without changing a line of application code.
Customer feedback
After the data is migrated to Lindorm, the data compression ratio is improved, and the hot data and the cold data are stored in different storage media. This significantly reduces our costs. Lindorm also provides higher performance and stability and lower write latency by using fewer nodes than open source HBase clusters. We have started to migrate the workloads of another business to Alibaba Cloud.
Challenges
An international marketing company ran its advertising platform on self-managed open source HBase clusters. Three problems limited growth:
-
Disaster recovery gaps: The clusters had poor disaster recovery capabilities. A single cluster failure could take down advertising features across nearly all clients, directly hitting company revenue.
-
High write volume: The platform received large numbers of read and write requests continuously. Daily incremental data reached up to 4 TB.
-
Cost-constrained retention: High compute and storage costs forced the team to retain data for only 7 days, leaving business requirements unmet.
Solution
| Challenge | How Lindorm addresses it |
|---|---|
| Poor disaster recovery | Lindorm is backed by Alibaba Cloud service level agreements (SLAs) and has been proven at scale across multiple critical services within Alibaba Group. The Lindorm expert team provides free technical support, freeing the marketing team from operations and maintenance (O&M). |
| Storage costs too high | Data compression reduces consumed storage space. Hot and cold data separation stores hot and cold data from the same table in different storage media — no changes to application code required. |
| Throughput bottlenecks under high write volume | The Batch Commit feature batches multiple requests in a single submission, improving performance by more than 3x and reducing response time. |
| 7-day data retention too short | Lower storage costs made it viable to extend the retention period to 4 times the original. |
The migration uses Lindorm's full compatibility with open source HBase, so no application code changes were needed.
Results
-
86% total cost reduction — total cost is less than 1/7 of the pre-migration cost.
-
4x longer data retention — from 7 days to 4 times the original, enabling business requirements that were previously unachievable.
-
Improved stability and performance — lower write latency with fewer nodes than the previous open source HBase clusters.
-
Zero application code changes — the migration preserved all existing application logic.
-
Reduced O&M burden — Alibaba Cloud manages operations, and the Lindorm expert team provides free technical support and scenario-specific performance tuning.