what's difference between pre-aggregation and post-aggregation ?
I have no experience in traditional BI and OLAP, not so clear to this
concept.

2016-05-10 13:51 GMT+08:00 ShaoFeng Shi <[email protected]>:

> This is easy in my mind; Firstly you need understand the "derived"
> dimension; "Derived" dimensions can only be from lookup table,  they will
> be derived from the pk/fk column at runtime; only the FK column will be
> built into cube; So if you use "derived", you don't need, and have no
> chance to declare them as hierarchy;
>
> Whether define them as a "derived" or "normal" + "hierarchy", depends on
> how you balance the pre-aggregation and post-aggregation.
>
>
>
>
> 2016-05-10 11:45 GMT+08:00 Mars J <[email protected]>:
>
>> It's very confused if I want to define a hierachy dim, I need to define
>> it be normal or derived and then add it to aggregation group for hierachy
>> dims. if this hierachy dims I want to define includes 3 columns, it will be
>> 3 dims if define it in normal dims or 1 dim if  define it in derived dims.
>> So how to handle it   ?
>>
>> 2016-05-10 10:53 GMT+08:00 Mars J <[email protected]>:
>>
>>> Hi,
>>>     There is something confused me .  In kylin 1.3 ,the dimension types
>>> are hierachy/derived/normal and mantory when create a cube in the
>>> dimensions step. In kylin 1.5.1 there 2 types including normal and derived
>>> in the dimensions step, and in the advanced setting , aggregation group has
>>> some aggregating manner of hierachy/mantory/joint.
>>>
>>>     I test 1.3  including 2 derived dims and 2 hierachy dims(also has 3
>>> columns each dims, level1-> level2 -> level3), I said it has 4 dims in the
>>> traditional BI way, but in kylin ,it's 10 dims.
>>>    In kylin 1.5.1, I can't create hierachy dims directly, but I can
>>> create it for derived dims or normal dims first, then in the aggregroup ,
>>> make it to hierachy dims. but I don't know how to count it in traditional
>>> BI way.
>>>
>>>    We can discuss about the dims, and the different dims definition.
>>>
>>
>>
>
>
> --
> Best regards,
>
> Shaofeng Shi
>
>

Reply via email to