10 responses

    • Me neither. This looks like something I really want to read. What are the chances of a kindle version?

  1. Pingback: New SSAS and MDX Training Course Dates « Chris Webb's BI Blog

    • Hey Chris,

      I’m a frequent follower and fan of your blog, and I just read the ‘Choosing the Right Model for Your Project’ section at the beginning of the book (which I just got a couple days ago).

      I’m currently in the unique position where I am starting an OLAP implementation with an organization that does not currently have one. My experience is in SSAS MD. So… I am truly at that fork in the road.

      Naturally the easier route for me is to choose MD because that is what I’m more familiar with — but, for the past year (at least), I have been reading the same writing on the wall as everyone else… this being that MS is spending the majority of it’s OLAP energy on tabular, not MD. I of course don’t want to paint anyone into a corner by choosing the wrong route.

      I was hoping you could possibly share your thoughts on the MD vs. Tabular question given some details about the project. Here is what I know right now after extracting a limited amount of data from the primary OLTP, transforming, and building a SSAS MD prototype.

      General stuff:

      Priority one is to offload the reporting strain that existing operational (SSRS reports) are putting on the OLTP source. The data analysis/exploration side of the sprectrum will be an on-going education process.

      I am leading the charge, and as mentioned, my experience is in MD

      I am the only one here with SSAS expertise, but the organization has several SQL pros that can/will come up to speed quickly.

      I have some experience with PowerPivot, so I have had a some exposure to tabular.

      I don’t get the sense that the organization has many (if any) Excel power users (candidate users that would ramp up quickly with DAX). In fact, just general pivot table slicing/dicing will be challenging initially…

      Specific Req’s/Features:

      I do have ragged hierarchies with fact data joining at all levels (eg: Insurance angencies that have agents and other angencies, that have their own agents under them…)

      I do have many to many relationships (eg: Insurance data with ‘Policies In Force’ at any give time, so in the prototype I’ve implemented a temporal snapshot fact table…)

      I have implemented my one date dimension as several role playing dimensions (for different date types)

      I have shown drill-through in a demo, and folks really liked it…

      That’s what I know right now…I’d appreciate any insight.

      Thanks so much,

      • Hmm, it sounds like you’ve got a tough choice. The ragged hierarchies won’t be properly ragged in Tabular, and the many-to-many relationships will complicate the code in all your measures, but there don’t seem to be any real deal-breakers as far as Tabular goes; so long as licensing isn’t an issue, I’d say you should go ahead and build a Tabular prototype to compare with your Multidimensional prototype. Only then are you going to be able to get a feel for all the pros/cons of each model for your project.

Leave a Reply to Ed Allison Cancel reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: