Organizations need both real-time system monitoring with fast response and historical data analysis for business insight. Traditional architectures typically require stitching together multiple systems to cover these two requirements. TDengine TSDB addresses this challenge with a unified architecture for time-series workloads.
Core design philosophy: a unified architecture built for time-series workloads
TDengine’s data model is the foundation that enables it to handle both real-time processing and deep analytics:
CREATE STABLE meters (ts TIMESTAMP, current FLOAT, voltage FLOAT, phase FLOAT) TAGS (location BINARY(64), groupId INT);
INSERT INTO meter_1 USING meters TAGS ("California.SanFrancisco", 1) VALUES (NOW, 10.3, 219, 0.32);
This design has three practical advantages. First, each device’s data is stored independently, avoiding lock contention and supporting high-concurrency writes. Second, the model supports fast single-device queries as well as cross-device aggregation. Third, the Tag system enables efficient filtering and grouping by metadata.
Real-time data processing capabilities
Continuous queries and Stream Processing
TDengine’s built-in continuous query feature processes data streams in real time. For example, it can calculate the average temperature when the ambient temperature exceeds 80 degrees for more than 10 minutes. Common use cases include real-time dashboard updates, anomaly detection, instant alerting, and live business-metric calculation.
Data Subscription (TMQ)
CREATE TOPIC meter_data AS SELECT * FROM meters WHERE ts >= NOW - 1h;
Data Subscription pushes real-time data to applications as it arrives. It supports multiple simultaneous consumers while maintaining delivery reliability.
Time-series data analytics capabilities
Time window analysis
TDengine provides a broad set of time window functions for complex time-series analysis. Sliding windows, for example, can compute a rolling average that advances every 5 minutes.
Multi-dimensional data analysis
Combined with the Tag system, TDengine supports complex multi-dimensional analysis. Users can group time-series data by geographic location and apply HAVING filters for refined aggregation.
Architectural optimization strategies
Multi-level data storage and management
TDengine uses an intelligent data tiering strategy. Data moves through four levels: real-time data in memory, hot data on SSD for recent time ranges, warm data on HDD for older history, and cold data in object storage for long-term archival. This approach keeps real-time data fast to access, stores historical data at low cost, and migrates data between tiers transparently to the application.
Distributed query optimization
In a cluster environment, the query optimizer selects the best execution plans. It runs cross-node queries in parallel and pushes computation down to data nodes where the data resides.
Real-world application scenarios
AIOps: continuous queries detect anomalies in real time, time window functions analyze performance trends, and the Tag system enables multi-dimensional root cause analysis.
Industrial IoT platforms: real-time Stream Processing monitors equipment status, long-term data storage supports predictive maintenance analysis, and the unified SQL interface simplifies application development.
Technical advantages summary
TDengine uses a single data model for both real-time processing and batch analytics, with no need to run separate systems. A built-in computation engine handles Stream Processing and complex analysis without additional components. The query optimizer adapts to different workload patterns without manual tuning. The architecture scales from edge devices to cloud clusters under one consistent model.
