How TDengine 3.0 Redefines Cloud‑Native Time‑Series Databases

TDengine Developer Conference Highlights

On August 13, Taos Data hosted the TDengine Developer Conference in Beijing, featuring founders and industry leaders who discussed open‑source trends, data‑architecture upgrades for IoT and IT operations, and announced the launch of TDengine 3.0.

Open‑Source Evolution and Community Impact

Since July 2019 TDengine’s core code has been open‑source, and its cluster version opened in August 2020. After five years the platform serves nearly 140,000 instances and over 100 enterprise customers, becoming a leading time‑series database in China.

Why Open‑Source Matters

Speakers emphasized that open‑source should be driven by altruism and long‑term vision, fostering collaboration and innovation rather than merely a branding or distribution tactic.

TDengine 3.0: Technical Breakthroughs

TDengine 3.0 introduces a cloud‑native, RAFT‑based distributed architecture that can handle up to one billion time series across 100 nodes, solves high‑cardinality challenges, and integrates native support for message queues, stream processing, and caching, simplifying system design.

The storage engine now supports multi‑engine hybrid storage and multi‑dimensional time‑series optimizations, building on the “one device‑one table” and “super table” concepts to deliver higher throughput and compression.

Additional features include flexible tag indexing, time‑range pre‑computation, schemaless ingestion, compatibility with Grafana, Google Data Studio, and robust backup, disaster‑recovery, and edge‑cloud collaboration capabilities. All source code is publicly available on GitHub.

Enterprise Use Cases

Companies such as SF Technology, OPPO, Yunda, Guance Cloud, and Kuayue have migrated to TDengine to address massive IoT and IT‑ops workloads. SF Technology reduced its monitoring platform from 21 servers to 3, cutting storage by 90% while improving write and query performance. OPPO selected TDengine for its high‑volume wearable data ingestion, citing low latency and easy integration. Yunda achieved 5,000 rows per second writes and sub‑second queries; Guance Cloud benefited from high reliability and multi‑tenant performance; Kuayue lowered daily disk growth from 22 GB to 1.4 GB, dramatically reducing operational costs.

Conclusion

TDengine 3.0 resolves the industry’s high‑cardinality problem and establishes a truly cloud‑native time‑series database, marking a revolutionary step for the field and positioning the platform for future digital‑era data‑architecture demands.