Forums  > Software  > any time-series database opensource projects with any inertia?  
Page 1 of 1
Display using:  


Total Posts: 44
Joined: Mar 2005
Posted: 2018-01-03 22:14
Anyone have opinions or awareness about opensource time-series DB's which have managed to achieve any sort of ongoing community & user base?

1) Standard disclaimers apply, and then some.


Total Posts: 1177
Joined: Feb 2007
Posted: 2018-01-06 07:28
Open source time series databases are 100% garbage. This is delicately amusing to me, as all you need to know to make a good one is, like some tiny subsection of libc. Wes could write one which is at least OK, but he probably won't as he's busy being Wes.

If you're cheap and can deal with writing code in J: J's proprietary database (jd) is very good. The user community is small, but Iverson and company are very responsive, and they'll cut you an eval license for cheap. I've been using it on and off for a few years; it's harder to use than Kx, but arguably more powerful.

"Learning, n. The kind of ignorance distinguishing the studious."


Total Posts: 1124
Joined: Jun 2007
Posted: 2018-01-06 12:23
You are talking about "pandas Wes" ?

Ich kam hierher und sah dich und deine Leute lächeln, und sagte mir: Maggette, scheiss auf den small talk, lass lieber deine Fäuste sprechen...


Total Posts: 368
Joined: Jan 2015
Posted: 2018-01-08 11:30
Without knowing your specific use case, I'm 95% certain that Postgres will work fine for you. Unless your data's huge or your queries are very bespoke, you probably don't need explicit time series support.

Seconding, @jslade, there isn't a good open source TSDB. Which is curious, because it really isn't that hard to design a good one from an architectural standpoint.

Good questions outrank easy answers. -Paul Samuelson


Total Posts: 18
Joined: Nov 2013
Posted: 2018-03-01 03:46
Not even InfluxDB? It has an open-source single node version


Total Posts: 264
Joined: Dec 2012
Posted: 2018-04-04 16:40
At a certain rocket company I've worked with, they had to roll their own as well. It's hard because there ends up being pretty extreme trades in design depending on the specifics of you consume the data. It is weird and frustrating that the open sources tools don't better embrace this and just focus.

There are no surprising facts, only models that are surprised by facts


Total Posts: 13
Joined: Jun 2011
Posted: 2018-04-05 15:42
I agree with @EspressoLover that Postgres should be fine most of the time.

InfluxDB and similar databases seem mostly focused on collection of basic metrics data (e.g. cpu usage) and I found them very limited on the query front, requiring a lot of pre-processing and denormalization before inserting into the DB or separate processors like Kapacitor to be able to do useful queries.

With Postgres you get the full SQL capabilities (joins, json, etc).

For large amounts of realtime data I've been looking around and TimescaleDB looks interesting.
It's an extension to Postgres that basically partitions your data over time and tags to keep the most recent data in memory and optimizes inserts for append-only data and you can query just like regular Postgres.
I haven't gotten around to actually trying it yet though.

Founding Member

Total Posts: 1359
Joined: Mar 2004
Posted: 2018-04-06 13:10
I've done some limited experimentation of the timescaleDB extension to postgres and it's been a good experience. Feels like normal postgres for all intents and purposes.

Capital Structure Demolition LLC Radiation


Total Posts: 202
Joined: Mar 2011
Posted: 2018-10-19 12:42
no guarantees but have a look at monetDBlite? I will have this thread send me an email if you respond here.
Previous Thread :: Next Thread 
Page 1 of 1