I am just writing a technical SOW before development begins for a client.
For part of our functionality around time capabilities we are impacted by whether a field is more rapidly accessible (such as by tags/indexed/partitioned columns).
All the glossary states for the timestamp portion of a point is: The date and time associated with a point. All time in InfluxDB is UTC.
Ideally we are looking for it to act as a tag value does: The value part of the key-value pair that makes up a tag. Tag values are strings and they store metadata. Tag values are indexed so queries on tag values are performant.
Time is a special case in time-series databases. The entire data-store is built around time.
Measurement + Tags make up metadata used to find the series. Tags are indexed.
Each series consists of points. Each point has time-stamp and one or more fields.
Points are ordered by time. Fields are not indexed.
In the queries time-stamp behaves more like a tag, but often needs a special syntax.
Time grouping/joining is based on intervals (windowing/tolerance), while tags are grouped by distinct values.
In the queries time-stamp behaves more like a tag, but often needs a special syntax.
Our system provides BI Visualizations and we have a feature for enabling a live visualization in realtime. One requirement is that the time field in question be indexed or partitioned. I just needed to make sure that the queries against time would fall under this category before development could begin as it is one our clients current pain points that needs resolution.
