As you may or may not have seen, a month or so ago the SQLCat team published a white paper on optimising the performance of ROLAP inside SSAS. You can download it here:
I wouldn’t normally bother commenting on something like this, not because it’s not worth reading, but because so much good stuff comes out of the SQLCat team you should be reading it all anyway! In summary, the paper shows some useful tricks to make ROLAP run faster and points out not only that ROLAP is the only option at data volumes of more than 5TB but (surprisingly) it can sometimes outperform MOLAP at smaller data volumes.
That’s not what I wanted to talk about though. While I was reading this paper it struck me that something that was not mentioned was Parallel Data Warehouse, which is going to be released relatively soon I assume. If you’ve got a lot of data (and cash) and you want to build a data warehouse on the Microsoft platform then you’re going to want to want to look at it, and if you’re investing this much in one area of the MS BI stack then it makes sense to invest in other areas like SSAS. And as you probably know ROLAP has always been a bit of a pain point for SSAS, so maybe the prospect of customers building ROLAP cubes on Parallel Data Warehouse has spurred the SQLCat team on to do some research in this area? Moving forward it would be good if the dev team spent some time optimising SSAS in ROLAP mode for both regular SQL Server and Parallel Data Warehouse because I suspect that there’s a lot of scope for optimisation. It will also be interesting to see where the dividing line is drawn for customers with large data volumes: should they be using ROLAP on Parallel Data Warehouse, or will the new super-scalable Vertipaq engine be able to handle multi-terabyte volumes even better?