On the server, Gemini leverages Analysis Services running as a service app in Sharepoint. The Gemini workbooks are stored as workbooks, but as needed the server can upload the embedded Gemini content to the AS server and provide typical services such as
refreshing data, or serving result sets to a client app as needed.
So, in a sense, yes the Gemini data is available as an SSAS cube - it's connection string is the URL of the xlsx file - and SSAS clients can query it. We show cool demos of Report Builder doing just that. But any other SSAS client could also do so. However,
what they see when they connect is a rather funky "cube." Technically, in this version, it doesn't have many of the regular cube constructs such as hierarchies, and for business scenarios a Gemini workbook will typically only address one business question,
or at least a relatively narrow domain of questions, compared to a typically broadly scoped OLAP solution.
However, you have indeed identified an important point. A Gemini workbook is not just a workbook - it is a real BI solution, and benefits from the openness of the MS BI platform. It's a bit like a Sagitttarius, which has the mind and smarts of a human, and
the powerful back end of a horse!
Right now I'm very focussed on turning our current dreams into reality, so I don't want to say too much about what happens "eventually."
But of course we are looking at all sort of interesting things we can do with this technology, and the community would be surprised if we were NOT considering some form of integration with the SQL Engine. But it's way too early to step down the path of what
that would look like, even it were feasible.
But the future for Gemini and the core technologies is surely something to watch!
We have done a lot of work on our mid-tier server to handle this scenario. In-memory applications posted to the server are loaded and unloaded as needed to serve the end-users demands. Of course, it is possible that a server could be overloaded - but the
server can be scaled out on a sharepoint farm, and we handle load-balancing between servers on the farm.
Do bear in mind that most Gemini apps will not be huge - very few will be running 100s of millions of rows - most will be using 10s or 100s of thousands of rows, and will be using Gemini for its integration, analytic and collaboration smarts.