Ask HN: Best database to store time series for analysis?

3 points by 0x008 ↗ HN
As the title says, which is the best database as a basis for a time series analysis system and why would you recommend it?

5 comments

[ 3.1 ms ] story [ 25.4 ms ] thread
There are several different types of timeseries workloads: financial, IoT/sensor data, monitoring metrics, timestamped events, etc.

Also do you need a custom datapoint schema?

Do you need a generic tags/labels/dimensions?

Do you need flexible datatypes, or numeric/float only?

Do you need roll-ups, aggregations, window functions, stream processing, etc.

Hm basically I am just looking for efficient storage and support for date type fields. The numeric values can be floats or integers.

Probably any SQL database will do the job, but I felt there was probably a more domain-specific solution maybe with a good pandas integration? I don’t really want to write my own SQL code and think much about schemas.

I'm a kdb+/q[1] developer. It's a proprietary timeseries database used in finance/trading, but also for IoT/sensor data.

But for your usecase I think TimescaleDB will be the best fit. It's an opensource extension for PostgreSQL, so you can use any PL/framework which can work with Postgres. And you have all the felxibility of RDBMS and can use standard SQL queries.

[1] https://en.wikipedia.org/wiki/Kdb%2B https://code.kx.com/q/

[2] https://www.timescale.com/

Thank you, I will check it out.

Do you have any experience with QuestDB?

No, just read couple of their blog posts.