LAIX uses Database Autonomy Service (DAS) to automate database O&M across a multi-engine stack, freeing its DBAs from reactive firefighting and reducing incident recovery time.
"DAS helps LAIX manage databases in a more professional and efficient manner. The AI-based optimized database engines provided by DAS help LAIX reduce a large amount of workloads for DBAs and make online troubleshooting more efficient. This way, database autonomy is realized." — LAIX
Challenges
As LAIX scaled its AI-powered English learning platform, its database team faced four compounding pressures:
Multi-engine complexity. LAIX runs ApsaraDB RDS, ApsaraDB for Redis, PolarDB, and ApsaraDB for HBase in parallel to support different workload types. Managing these databases centrally — with consistent visibility and control — was difficult.
Pace of business iteration. Rapid product releases created sustained pressure on both development and O&M teams. DBAs could no longer rely on experience alone to keep pace with the rate of change.
Service continuity expectations. Growing user demand raised the bar for availability. When production incidents occurred, recovery time was unpredictable, creating risk that was difficult to bound.
Cost efficiency. LAIX aimed to build cost-effective infrastructure by leveraging the scalability of cloud-native database services without proportionally growing the operations team.
Solution
LAIX adopted DAS as the foundation for intelligent database O&M, enabling autonomy across its multi-engine environment.
Autonomous operations — LAIX activated DAS's core automation features across its database fleet:
24/7 anomaly detection catches issues continuously, using machine learning to identify problems earlier than threshold-based monitoring.
SQL diagnosis enables full request analysis to identify and diagnose query-level performance issues.
Automatic SQL throttling triggers when configured conditions are met, protecting databases from traffic spikes without manual intervention.
Automatic scaling for performance optimization scales instances up during peak load and back down as CPU utilization drops, keeping resources matched to actual demand.
Automatic SQL optimization creates and removes indexes using online DDL, avoiding table locks while keeping query performance tuned to evolving workloads.
Intelligent stress testing validates whether instance storage capacity meets business requirements before going live with new workloads.
API-driven incident management — LAIX integrated the DAS API into its operations workflow to automatically troubleshoot production exceptions, run intelligent diagnosis-based database optimization, and perform digitalized assessment of database capacity. This transformed incident response from ad hoc escalation into a structured, repeatable process. LAIX also developed a methodology for intelligent database O&M based on these capabilities.
Shift in DBA focus — With routine O&M automated, LAIX's DBAs now concentrate on business-oriented data architecture design rather than operational maintenance — a structural change in how the team creates value.
Results
Reduced DBA workload. DAS handles the monitoring, diagnosis, and remediation tasks that previously consumed DBA time.
Shorter incident recovery time. Automated detection and SQL throttling reduced the average time required to restore service after a failure.
Validated capacity before scale. Intelligent stress testing gives LAIX confidence that storage capacity meets business requirements ahead of peak events.
About LAIX
Founded in September 2012, LAIX is a science and technology-driven education company focused on intelligent English learning. LAIX has built a large-scale English phonetic database of Chinese learners — capturing approximately 3.9 billion minutes of conversations and 52.4 billion sentences as of December 31, 2020. On top of this dataset, LAIX developed an engine for assessing oral English proficiency and writing proficiency and a self-adaptive learning system to support the full range of English language skills: listening, speaking, reading, and writing.