Ok I am going to blog this Channel 9 Video interview of the Gemini team as I am watching it so I may ask and answer my own questions!
Firstly the people being interviewed are from the Analysis Services team. Now this has huge implications as it means that AS and cube based people are developing this product. I have always felt (but not proved) that a BIG part of the problems that arose when connecting Excel to Analysis Services Cubes was the disconnect between the history, background and goals to 2 very different teams. That Gemini comes from the AS side bodes well.
They have achieved the scalability to allow for millions of rows by looking at, not the rows but the columns and compressing on columns. So a Post/Zip code column can be significantly compressed and a product code probably even more.
The compression happens when the data is loaded.
They nicely differentiate between “business” and pure analytics. This is for Business use where a star schema really works.
So it seems that a big chunk of what Gemini is trying to do under the hood is normalising a very flat de-normalised structure into a normalised one in memory.
They are trying to bridge the gap between the users who are familiar with BI and the traditional Excel users. I note that there is no mention of the users familiar with the Line of Business apps Gemini could be hooked up to.
“How do you take that (BI) technology to a mass market” My concern is that this is not and never has been an end user tool issue but a source data issue. There are tools out there but the data is not out there!
Classic quote from Charles (The interviewer) “Excel users don’t realise they are programming” I would say they are not programming as much as data modelling. Just as dangerous!
It seems that the team do realise the danger areas they are heading into when they discuss the implications of formatting –V – data types and relationships and getting Excel users to understand that “that’s pretty demanding” – No kidding!
Classic balance between the Clippy UI and the power UI.
12:33 (just spotted the time marker) There is reference to the internal politics that I mentioned in my first paragraph and more importantly how they blended the team to overcome them.
13:42 The AS team also embedded people in the Excel team and they worked on the new Slicers feature. Lets hope the resulting MDX queries perform properly!
17:08 Mashing up data. This still scares me. The ability to let business users hack multiple data sets together and most importantly make decisions on the result sends shivers down my spine!
19:20 There is a reason that the data warehouse does not have the 30 lines from Gartner. It is because there is no valid and consistent mechanism for matching the measures they apply to the measures available in the DW! Apples meet Pears!
19:54 Ok they realise that data wont come directly from back end systems but will be derived from reports from those systems. So if the reports are aggregating and managing the data what is the gain from being able to handle large data sets? Why not just write a SSRS import tool or use the existing export functions in SSRS?
23:11 Why is Gemini called Gemini? Bridging the gap between Information Workers and IT
This without doubt a very cool and smart toolkit for obtaining, mapping and displaying data. However I still have significant concerns about the result of giving many business users the impression that they can throw this tool at internal LoB solutions and external data and make decisions based on the results. On the positive side I see this as another boost to the concept of the corporate data warehouse/cube and a further drive to LoB suppliers to provide a proper structured data interface to their products.
Now if only Microsoft had held faith with PerformancePoint then they could have built a killer suite of analysis and management tools. Maybe one day.