[Mondrian] <join>,<Query> and <view> for snowflake schemas
isaias.sanchez.l at gmail.com
Mon Mar 21 07:21:10 EDT 2016
I've tested all and all three options perform more or less the same. The
difference can be on memory available for database cache or the one
available for mondrian cache, also if you want to use some database
tricks for the join is preferable to use <View> or <Query> (these two
are the same in performance in my opinion).
Talking about cache probably you have some dimension table with 10
columns and some unused or expired rows. If you use <Join> mondrian will
take all that data even if it won't be used in the cube. With a View or
a Query you can filter that.
As a test I recommend you create your Query and check saiku or database
logs to check the Query created by <Join> and compare their performance
against each other, there you will have the better option.
Selina Tech wrote on 19/03/16 00:32:
> Dear All:
> When I study the Mondrian schema. I think we have three different
> ways for the Dimension of snowflake schema.
> Way 1: Use <Join> as the example at
> Way 2: Use <Query> to define a ‘table’, and use this ‘table’ for
> Dimension table
> Way 3. Use <View> to define a view ‘table’ by <SQL> and use this
> ‘table’ for Dimension table
> Is any one of them wrong? if they are right, how could I choose
> one of them? How is the different on performance? when we use <view>,
> is this <view> will generated in memory, disk or database? what happen
> if the result of view is huge on size?
> Your any help is highly appreciated.
> Mondrian mailing list
> Mondrian at pentaho.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Mondrian