March 12, 2015 at 9:58 am
I'm new to SSAS and I'm having a meltdown because I can't really get my head round dimensions and hierarchies.
I've worked through a few tutorials which use AdventureWorks and all these perfect examples which bear little or no resemblance to real life. I've built a robust looking Calendar dimension but that's easy because its all so well structured. Every date belongs to a particular week and every week belongs to a particular month and every month belongs to a particular quarter etc. January 1st is always going to be a member of week 1, month 1, quarter 1.
But here's the rub, real life isn't as predictable and stable as a calendar.
Lets say you have an table containing personal information. You might have gender and marital status in there and
even though they both relate to one person, they're unrelated to each other so how do you put those in a hierarchy? Which one goes first and which second because you can get combinations of either. Add in further similar factors and the thing starts getting really messy.
I'd really do with some pointers. I suspect I just need someone to explain it properly and I'll be well away.
Thanks.
March 12, 2015 at 9:31 pm
I would recommend that you create hierarchies and relationships between attributes where there is actually a relationship between the attributes. This allows SSAS to utilise the aggregations for each level of the hierarchy in its parent which is how you get good performance from SSAS.
This does not stop users from including Gender and Marital Status in a report.
Yes, I have seen people define hierarchies using attributes that are unrelated. I am yet to find a good reason to do it despite the protestations by the designers of those hierarchies.
Viewing 2 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply