Is it the end of the Relational Dimensional Data Warehouse ?

With all the technologies of cubes, columnar databases, big data, advanced semantic layers, self-service BI and tools like Power Pivot and Power View from Microsoft for example, is it still relevant to build a traditional relational data warehouse? A Kimball like with facts, dimensions (surrogate keys, Slow Changing Dimensions and etc…)

Don't believe the marketing hype. Dimensional warehouses will be around for years to come.

Cubes have limited capacity.

Columnar databases are relational.

Structured big data is handled in relational environments (MPP systems). Unstructured data is sometimes. You should remember that the whole point of Hadoop is to address MPP for those who don't have an MPP database system. It is not a replacement, and a very poor substitute when dealing with structured data. It is also very inefficient both in personnel and hardware.

Semantic layers (and 'self service' tools) have, and will continue to coexist with dimensional DW implementations. They are complementary technologies.

It's definitely not the end.

In-memory solutions actually work better against a star schema than a snow-flake or 3NF solution.
Storing your entire data warehouse in-memory is also not sustainable for large (multi-terabyte) data warehouses.

In-memory has it's place, but obviously has a cost in terms of hardware. Yahoo have a 24Tb SSAS cube (last I heard, probably bigger now) - imagine how expensive that would be to put into memory!