Categories
Database Real-Time Analytics Edge Computing Industrial IoT Data Storage Time Series Database IoT DatabaseStars
Forks
Watchers
Developer links
Apache IoTDB
Sensors on factory floors, wind turbines, and connected vehicles produce readings every second, and Apache IoTDB is the time-series database built to store those readings and answer questions over them. Devices write through client sessions that tolerate out-of-order, delayed, and multi-frequency arrivals, so a gateway that reconnects after an outage can backfill history without corrupting the timeline. Series live in a device hierarchy you can browse like a directory tree, or in tables queried with standard SQL, common table expressions, set operations, and more than one hundred built-in aggregation and time-series functions. The TsFile columnar format shares one time column across aligned series and applies encodings designed for sensor data, which keeps high-frequency archives compact while Spark and Hadoop jobs read the same files directly for larger analysis. Pipe synchronization streams records or transfers whole TsFiles between edge sites and central clusters with path and device filtering, so plants keep collecting locally while headquarters queries a continuous copy. AINode brings forecasting, anomaly detection, and missing-value imputation into the database itself: register a model with SQL, then ask for predictions in the same query language instead of shipping measurements to a separate machine learning service. A low-memory Edge edition runs collection on constrained gateways, and clusters scale by adding DataNodes under ConfigNode coordination with automatic load balancing. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.
Benefits
- Built For Sensor Data
- TsFile columnar storage shares one time column across aligned series and applies time-series encodings, so high-frequency device readings compress tightly while remaining directly queryable without ETL into another analytical store.
- Edge To Cloud Sync
- Pipe streams records or transfers TsFiles between edge and cloud nodes with path and device filtering, so gateways keep collecting locally and headquarters still receives a continuous, synchronized copy.
- Forecasting Inside The Database
- AINode registers time-series models and runs forecasting, anomaly detection, and imputation through SQL statements, so predictions happen where the measurements already live instead of on a separate machine learning service.
- Scales Across Distributed Clusters
- ConfigNodes coordinate metadata while DataNodes store and query data in a distributed architecture with high availability, automatic load balancing, and support for heterogeneous clusters that grow node by node.
Features
- TsFile Columnar Storage
- Apache TsFile stores time and value columns with encodings designed for regular sensor intervals, and Spark or Hadoop jobs can read the same files directly.
- Tree And Table Models
- Organize series in a device hierarchy with the tree model, or use the table model with standard SQL including CTEs, set operations, and over one hundred built-in functions.
- High-Speed Ingestion
- Clients for Java, Python, C, and Node.js write through sessions that tolerate out-of-order, delayed, and multi-frequency data arriving from millions of devices at once.
- Pipe Data Synchronization
- Pipe synchronizes data between IoTDB instances at record level or as TsFiles, with filtering by path or device for edge, cloud, and multi-site replication topologies.
- AINode SQL Inference
- CREATE MODEL registers forecasting and anomaly detection models, and SQL forecast calls return predictions without moving data to an external machine learning platform.
