I have been keeping a close eye on machinations in the SQL Server Data Services (SSDS) arena and whilst the product hasn't reached the masses yet (I'm still waiting for my beta invite - hint hint) there is a still a lot of activity going on. There is a very active blog called The Long Term Storecast (http://blogs.msdn.com/ssds/default.aspx) for example and Technical Fellow Dave Campbell gave a very eloquent demo incorporating SSDS during Bill Gates' keynote from TechEd yesterday.
I've been in future-gazing mode again (as I am wont to do) and have been wondering about what features SSDS may incorporate further down the line. The SSDS team have talked candidly about how they will look to leverage the full feature set of SQL Server in the future so I suspect that one day we will see such things as a cloud-based queuing service (i.e. SQL Service Broker), cloud-based procedural logic (i.e. stored procedures) and persisted cloud-based query definitions (i.e. views).
As a BI guy first and foremost though I am wondering whether SSDS will incorporate a cloud-based aggregation engine; in other words, cloud-based Analysis Services. Scalable though I'm sure SSDS will be there is bound to be latency if the query engine is attempting to aggregate, say, sales figures per region when that data is dispersed across multiple servers and perhaps even multiple data centres. Analysis Services has proven that pre-calculating aggregated data can be hugely beneficial for large data volumes so I don't think its beyond the realms of possibility that we will get data aggregation in SSDS someday.
In an ideal world the delivery of pre-calculated aggregations would be transparent to us. In other words, I issue a query and the query engine makes the decision as to whether to use the base data or pre-calculated aggregations, I as a user don't actually care how that data is stored and I CERTAINLY don't want to have to decide:
In an on-premise solution today we have to consider both because we have to choose between SQL or Analysis Services, T-SQL or MDX. SSDS provides an abstraction layer that may mean we don't have to make this decision; I really hope that proves to be the case.
[Of course, this raises the question as to why in SQL Server today we have two seperate query engines for this stuff rather than having Analysis Services baked into the database engine but that's a political hot potato that I'm not going to tackle right now.]