A game company migrated its advertising business data from a self-managed HBase hybrid cluster to Lindorm, reducing average response time from 500–800 milliseconds to 20 milliseconds — a performance improvement of at least 30 times.
Challenges
Unstable performance from resource contention: The self-managed cluster ran multiple big data services on shared infrastructure. Services competed for resources, causing unpredictable request latency and poor stability.
High infrastructure costs: The cluster required more than thirty 32-core 128 GB Elastic Compute Service (ECS) instances, with data exceeding 24 TB and storage overhead continuing to rise.
Limited disaster recovery: A single cluster served all advertising workloads with no cross-zone failover capability, leaving the business exposed to outages.
How Lindorm addressed each challenge
| Challenge | How Lindorm addresses it |
|---|---|
| Unstable performance from resource contention | Migrating to an independent Lindorm cluster eliminated cross-service interference. The garbage collection (GC) optimization feature and the triplicate architecture of Lindorm Log Consensus (LLC) reduced maximum response latency for 99.9% of requests. |
| High infrastructure costs | The Lindorm expert team suggested enabling the data compression feature, achieving a compression ratio of up to 13:1 and reducing storage overhead to less than 5% of what HBase required. |
| Limited disaster recovery | Primary and secondary clusters are deployed in the same region, enabling immediate workload failover when the primary cluster fails. This provides cross-zone disaster recovery without additional infrastructure. |
Results
At least 30x performance improvement — average response time dropped from 500–800 milliseconds to 20 milliseconds, directly increasing advertising system throughput.
Storage overhead reduced to under 5% of the pre-migration footprint — data compression at a 13:1 ratio made the rising storage costs sustainable.
Cross-zone disaster recovery — primary and secondary clusters in the same region ensure business continuity even in extreme failure scenarios.
Freed from O&M — Alibaba Cloud provides service level agreement (SLA) guarantees, a full suite of data management services, and a dedicated Lindorm expert team for free technical support.
"The performance of our self-managed cluster was not stable and we could not obtain fixed values to evaluate the performance. After the data was migrated to Lindorm, the system stability was significantly improved and the performance was improved by at least 30 times. The system still has room for stability improvement and performance improvement."