We are currently evaluating columnar databases for our application.Below is an abstract description of our requirement.
Our app is like a dashboard powering and drilling down among several tables having wide columns
Our tables have around 1000 columns called custom fields.
There are currently 20 such fixed tables along with 30 other tables.
Data types are a mixture of int,bigint,tinyint,double,decimal and text.
Text columns form the majority with 30% of the columns i.e 300 columns approx.
This is like a dashboard view comprising of 20 columns which can be mixed/joined among the 50 tables i.e the view can be created with columns comprising from any of the tables.There can be joins,where conditions when defining the view itself.
The user can sort by any of the columns in the view in any manner i.e ascending/descending.
There can be where conditions in any of the columns and the view.
Text columns can have conditions like starts with (abc%), ends with (%abc), contains (%abc%).They can also be sorted based on alphabetical ordering.
Integer columns can have aggregate functions such as sum,>,<,>=<= etc.
The user can join these tables based on the unique/primary key column.Considering the max amount, there can maximum 50 joins although this is not the general case.
We are currently analyzing MonetDB, but not much confident about its stability.
Is there any person who can confirm this.I can even cross post this in the developers group as well if needed.
Please get back to me in case of any questions.
Thanks for the wonderful work you are already doing.