Open Source alternative to kdb+

[acatalano]

Is anyone familiar with an open source alternative to the kdb+ time series database?

 

[alain]

I played with K once around 2000. At that time, there was no "free-lunch". I remember the company that produced KDB used to offer a limited version for free that you could play with but I haven't seen anything like it in the open source world.

The only thing I remember about K is that it was superfast but cryptic (APL based I think).

 

[acatalano]

Hi Alain-

Yes I played around with the limited trial version as well. It is very fast but also very expensive.

Thanks for the reply back.

-regards

 

[Wallstyouth]

Is anyone familiar with an open source alternative to the kdb+ time series database?

Opensource DB on steriods MonetDB - http://monetdb.cwi.nl/ not very good for time series data as k but its much better than most of the opensource stuff available today

Checkout HDF5 its opensource but API only - http://hdf.ncsa.uiuc.edu/whatishdf5.html it works very well with Esper Esper -

Cheaper than kdb but comes very close in speed Time Series Database - Xenomorph Database (XDB)
http://monetdb.cwi.nl/

 

[rholowczak]

Not open source but may be less expensive...

There seem to be many new entrants into this space.
Vhayu, StreamBase and others. I think StreamBase offers a free demo version.
You might also look at Times Ten (acquired by Oracle - also has a trial version):
http://www.oracle.com/technology/products/timesten/index.html


Cheers,

Prof .H.

 

[eheiorjgeio]

Hey,

I thought I would give this thread a bump as quite a while has gone by, I was wondering whether there were any new kdb+ like open source alternatives that have a similar feature set, or performance.

I've recently installed and worked with a open source trading platform tool called marketcetera which looks promising, but it lacks the historical data features currently.

Thanks guys,

T,

 

[MerlinBrasil]

Updates?

Hi,

Been quite a while since anyone posted to this thread.

Anybody have any new comments or suggestions to add?

Thanks,
Merlin

 

[Yike Lu]

If you don't need it for production, (i.e., academic or other non-commercial use) there's a 32 bit trial version.

 

[Lyosha]

trial version cuts you off at a certain point though, which I would imagine is thoroughly unsatisfying...

 

[kirylm]

trial version cuts you off at a certain point though, which I would imagine is thoroughly unsatisfying...

I could not find any pricing on kbd+. They all refer to sales for a quote. What is the price range (if anyone know concrete numbers?)

 

[DominiConnor]

I've been talking to Xenomorph, who produce the DB rmentioned by WallStYouth.

The idea we've been kicking around is to provide a "student edition" that's cheap / free, possibly populated with some time series.

Although I think that would be great for some students, I haven't yet been able to work out whether there would be enough takers to make this a firm proposition.

 

[Eugene Krel]

I could not find any pricing on kbd+. They all refer to sales for a quote. What is the price range (if anyone know concrete numbers?)

expensive

 

[immortal]

Seriously? Anyone?

I am not sure if there is enough people who really realize that, before any serious quant work could be done, a time series database is a must.

I want to see if there is really an interest on this forum to:

1: Develop a time series data storage
2: Develop data collectors (data providers) that populate the database
3: Share the data.

and what kind of time/money commitment each interested person want to contribute

This is not a "call." It is more of a survey. Please feel free to chime in..

thanks!

 

[kimosabe]

Is anyone familiar with an open source alternative to the kdb+ time series database?

There is always A+. It was a predecessor to kdb.

 

[mysterion]

Maybe try http://widebase.github.com/

 

[Dave Haan]

There is always A+. It was a predecessor to kdb.

A+ is an C-APL hybrid. K is a crunchy Lisp kernel with APL sugarcoating. (I was expert in both, and APL of course ...)
and Q is K/kdb+ with SQL brought onboard (after my time) (sorry for delay I've been away).

 

[Yike Lu]

Open source K:
https://github.com/kevinlawler/kona/

Widebase looks interesting - you are the creator and maintainer, mysterion? Do everybody a favor and don't implement strict SQL (the COBOL of query languages) as the main query language. The q model isn't bad: do embed sql operations inside your base language.

 

[mysterion]

Open source K:
https://github.com/kevinlawler/kona/

Widebase looks interesting - you are the creator and maintainer, mysterion? Do everybody a favor and don't implement strict SQL (the COBOL of query languages) as the main query language. The q model isn't bad: do embed sql operations inside your base language.

Yep, I'm the creator and maintainer. I tried first the Apache Cassandra database, based on the tutorial Basic Time Series with Cassandra: http://rubyscale.com/blog/2011/03/06/basic-time-series-with-cassandra/. But, this was not the right approach for my case. In the end, I have written own database for large time series and familiar data model like KDB+.

 

[Yike Lu]

Interesting, I will take a look at widebase when I get a chance then. Thanks for sharing!

 

[mysterion]

Try the new features like charting and UI tables by 0.3.4-SNAPSHOT