High number of fields per measurement - can it cause storage overhead?


What is the overhead of having high number of fields (200 fields of type float) per single measurement ?

What is the impact on tsi storage ? Can storage space occupation explode ?
What is the impact on compaction performance ?

Context: Currently I’m using telegraf elasticsearch input plugin to collect stats from multiple large ES clusters,
and I noticed inflxudb database disk size grows up pretty quickly (300GB for 6 months), and ful compactions are taking lots of CPU/IO/memory resources. Therefore I wonder whether high number of filelds can cause this.

Hello @Anatoliy_Dobrosynets,
Fields aren’t indexed so having multiple fields and no tags, or filtering for fields, will decrease query performance. Since they’re not indexed they should decrease the impact on tsi storage.

However, I’ll share your question with the influxdb team. Thanks for your patience.