Time series storage

Hi all,
help me out please to understand where would we store this type of data…

Where do we put time series data - in memory or on disk?
Say a simple one like in the examples on training (stock, time, value)…

Then, if we also want to push time series through a transform pipeline say daily, not streaming live, wouldn’t it be faster if it has to write to a in-memory table?

On the other hand, the tables to write to would get large quickly meaning RAM requirements would rise too.

Thanks!

Hi Nikola, welcome to the SingleStore forums!

Most of our users who have large time series data sets use columnstore tables to store their series, so they are stored on disk.

Check out this documentation topic, which addresses your question.

Also, take a look at FIRST, LAST, TIME_BUCKET, and the SERIES TIMESTAMP DDL tag.

Thanks for the reply!