infrastructure//storage//time-series database

A time-series database is a database designed for data indexed by time: measurements with a timestamp, tags or labels that describe the source, and the fields that carry values. InfluxDB and TimescaleDB are the usual examples, and in a plant they are the open alternatives to a commercial historian such as PI; the model appears everywhere from server observability to IIoT. It is where the raw time series lives before anyone models it as a signal, and its labels and retention policy decide which questions a model can even be asked.

A time-series database is a database designed for data indexed by time: measurements with a timestamp, tags or labels that describe the source, and the fields that carry values. InfluxDB and TimescaleDB are the usual examples, and in a plant they are the open alternatives to a commercial historian such as PI; the model appears everywhere from server observability to IIoT. It is where the raw time series lives before anyone models it as a signal, and its labels and retention policy decide which questions a model can even be asked.

InfluxDB is built for time series from the ground up. A relational managed database optimizes rows and transactions; a time-series store optimizes append-only writes and range scans over time. TimescaleDB is the attempt to have both in one engine: it extends PostgreSQL, keeping the whole relational and SQL ecosystem while storing series efficiently. That second design answers a real need, because asking for the vibration of every pump from manufacturer X that was producing product Y when a fault occurred is a time-series question joined with relational context (data contextualization).

The database stores whatever timestamp it is handed, and it cannot tell which one that was. If the gateway stamps a sample when it arrives instead of when the sensor took it, every downstream analysis inherits the network's latency and jitter as if they were part of the process; the stamp has to be taken at acquisition and carried with the value (time synchronization).

In plants these databases coexist with the older industrial historian rather than replacing it: historian, data lake or cloud, TSDB and SQL each keep a role.