Lindorm Stream is a serverless, SQL-native real-time data processing engine that uses Flink as its unified compute engine. It handles data ingestion, transformation, and analysis for Internet of Vehicles (IoV), Internet of Things (IoT), and internet workloads — so you write SQL, not infrastructure code.
Core capabilities
Lindorm Stream is designed around three principles: serverless operation, standard SQL access, and operational completeness.
| Capability | What it does |
|---|---|
| Serverless | Runs on Alibaba Cloud Elastic Container Instance (ECI) containers. No servers to provision or manage — deploy a job and start processing immediately. |
| Auto scaling | Scales compute up and down based on time-of-day rules for peak and off-peak periods, and on real-time cluster load to optimize resource utilization. |
| Resource group isolation | Physically separates multiple resource groups so stream jobs stay independent and secure from each other. |
| Standard SQL access | Write data pipelines using standard SQL and real-time extract, transform, and load (ETL) SQL. |
| Built-in connectors | Native connectors for LindormTable, Lindorm Column, and Lindorm Search share metadata across engines and preserve data consistency, reducing integration work. |
| Database CDC | Integrates Flink Change Data Capture (CDC) to capture data changes from MySQL, Oracle, and other databases in real time. Use this to build heterogeneous data synchronization pipelines without custom connectors. |
| Unified stream and batch processing | Runs streaming and batch jobs on the same Flink engine. One technology stack handles both processing modes. |
| Stream O&M platform | An integrated platform covering the full job lifecycle: development, deployment, operations and maintenance (O&M), monitoring, and alerting. |
Use cases
Real-time ingestion
Use Lindorm Stream to decode, parse, and ingest high-velocity device data as it arrives. In a typical IoV deployment, the engine transforms raw in-vehicle telemetry into structured records before writing them to storage.
Real-time ETL
Run pre-aggregation and pre-JOIN operations on table data continuously, writing results to a sink table as they compute. This works like a real-time materialized view: downstream queries always read pre-computed results rather than scanning raw data, which reduces query latency for reporting workloads.
Real-time analysis
Run continuous analysis on Lindorm CDC data for use cases such as anomaly detection in sensor streams, geo-fencing for fleet management, reporting dashboards that reflect current state, and Complex Event Processing (CEP) for pattern matching across event sequences.
