I hope you were able to attend my free webinar on Choosing the Right Analysis Services: MOLAP vs. Tabular on February 11, 2014 that I delivered with Dustin Ryan. If you weren’t you can now watch the recording here.
Traditionally following a webinar I try to post a quick Q&A session for everything I was not able to answer in the allotted time.
Q: Since Tabular solutions are many ways better than Multidimensional..then my question is when to go for Multidimensional solution
You would likely still choose Multidimensional for things that Tabular does not have full feature parity. For example, Tabular does not have the ability do writeback.
Some still choose Multidimensional because their data model is very complex and Multidimensional is more forgiving in this regard. Last, Multidimensional is a more scalable for larger datasets because it is less (still uses memory, CPU, etc.. excessively) reliant on pure memory to get the performance you need, unlike Tabular. As we mentioned in the webinar, Yahoo chose Multidimensional because it would better scale their 26 terabyte cube.
Q: When deciding whether to use Tabular or Multidimensional, do you have to have SSAS installed / set up specifically for one or the other?
Yes, when setting up SSAS Multidimensional or Tabular they are separate installs. Meaning they can’t be installed at the same time with the SQL Server installer. Because of their resource demands you traditionally want to install these instances on separate servers if possible.
Q: How do i link if column have more than one column is key column in tabular?
Great question. Usually, if you have tables that need to be joined together on multiple columns you have create a key column that combine the values into a single column either through DAX or through the source query extract.