Well, admittedly I'm old school here when it comes to this topic. I have to think that we would be smart to keep our servers focused on the traditional tasks of storage and retrieval of DATA for presentation to the front-end (either user machines or other servers) for them to analyze and create the INFORMATION for the end user.  While a single or a few analytical processes may be OK, unless there is really, really good design and development, it is very likely that there are going to be waits while data is manipulated taking cycles away from the data movement.  And unless we now get DBA's highly trained and experienced in the analysis process, our main focus is going to suffer.  As the old idea said, that I think originated years ago even before the client/server concept existed, use the KISS method - keep it simple, Stupid.  And of course, this harks back to earlier discussions regarding the continuing issue of how much detail data to retain online, how much summarization to do, etc.  Combining ever growing quantities of data with ever-increasing processing of that data appears to be a place we may not want to go.  When it comes to taking on all the processing tasks, it seems maybe we should 'be careful what we ask for - we just might get it'.