Does Microsoft OLAP cube dead? Well, the internet is full of technologies and products that blogs writers thought are dead. Just Google 'is dead' technology and you will see a full list, Including Paul McCartney is dead.
But is seem that Microsoft is close to kill the OLAP cubes as we know them. There is new king on town: In-Memory BI. Few years ago Qlick-Tech start to sell the QlikView product with the slogan: 'Quarter develop time'. I didn't think then and I don't think today that you can count on an in-memory product that will replace your enterprise BI system that is built on an ETL product, data warehouse and a reporter tool. The main reason in my opinion was the need for large database and a good and quick tool for data cleansing.
Those days the hardware is much stronger. A server built with 8 CPU with 16 GB RAM 64BIT is now standard. If you have such a large RAM and quick CPU's, why bather to take the data with ETL, store it on data warehouse and build cubes on top of it?
Let's just pull all data from the production server into the BI server RAM and then let the users query it.
In my opinion that is the logic behind the PowerPivot at Excel 2010 and now the Tabular model on SSAS 2011 (Denali).
When you install SQL Server 2011 Analysis Services you have to decide rather this server will be used for the regular OLAP cubes or the new kids in town Tabular model. If you need both Tabular and OLAP cubes then you should install two instances.
Tabular model can work in two different ways. Offline (Verti-pack cache) and online mode. On the offline mode the system work like OLAP cube: first use a process to pull data from the data source into memory (Verti-pack cache) and then the user make queries against the SSAS cache. On the online mode the user queries run through the SSAS tabular model into the source data for each query.
I do remember the promise of SQL Server 2005: the DSV will make the data source transparent to the business user. Well that's didn't append. On top of the DSV you had to build a cube which was later explored by the user.
But now on SQL Server 2011, the promise is start to get alive. While the user is exploring the tabular model he or she can actually querying the production database, the data warehouse or a CSV file. They don't know it unless told by the developer because data source is actually at the end transparent to them as promised on 2005.
At the end, I believe that the traditional OLAP cubes with MDX will remain for one of the two cases: very large database or cubes with very complicated calculations.