A time-series database decision starts with your actual ingest and query patterns, not a benchmark headline. TDengine targets timestamped data, but a useful pilot must reflect your devices, tags, retention period, and operational constraints.
Run a representative pilot
- Load a sanitized sample containing normal traffic, gaps, late events, and high-cardinality tags. Measure ingest throughput and the queries people actually use: recent readings, aggregates, and comparisons across devices.
- Define retention before loading production data. TDengine's database
KEEP setting can remove expired time partitions; verify that its behavior matches legal and recovery requirements.
- Test schema changes, replication, alerting integrations, backup, restore, and export. A fast dashboard is not enough if recovery is uncertain.
- Compare infrastructure and operator time with the database you already run. Keep a rollback path until the new system survives both a load spike and a restore drill.
The TDengine architecture guide explains the time-based storage model. Record measured results and unresolved risks beside each requirement.
Akousa