I attended this years  PASS Summit in Seattle and it turned out to be a great event.  Many of the sessions I attended were standing room only.  During the first Keynote, the presenter showed off a few new features in Kilimanjaro.  Kilimanjaro is the code name for the next release of Microsoft SQL Server and scheduled for release the first half of 2010.  This release will have heavy focus on BI functionality.

As you may recall, earlier this summer Microsoft purchased DATAAllegro which was well known for their data warehouse appliance.  Well, in Kilimanjaro, Microsoft is incorporating several enhancements based off of the DATAAllegro technologies into SQL Server.  The project code name for this is Madison.  With this initiative, they will be targeting data warehouses greater than 1TB.

The other major feature is code named Project Gemini which is an all in memory extension to Analysis Services - at least that is my understanding.  The demo they showed was incredibly impressive.  They created a model with 10 million rows in excel and were sorting and pivoting in fractions of a second...all running on a laptop.  This model could be saved to a SharePoint site and therefore accessed quickly and easily by other team members.  I didn't hear anything mentioned with regards to write back being implemented with Gemini but will be very interesting to see if it is.

You will see this or a similar code sample when referring to clearing the analysis services cache.  However, I have yet to find a good resource that explains exactly what running an empty MDX query against a cube really is doing. 

Here is an example of code to clear the cache:

<Batch xmlns="http://schemas.microsoft.com/analysisservices/2003/engine">

    <ClearCache>

        <Object>

            <DatabaseID>Adventure Works DW</DatabaseID>

        </Object>

    </ClearCache>

</Batch>

GO

SELECT {} ON 0

FROM [Adventure Works]

From what I have gathered (mainly due to feedback received from Darren Gosbell,http://geekswithblogs.net/darrengosbell/Default.aspx) is that running the empty MDX query on a cube essentially reloads in cache all of the calculated measures, named sets, scoping, etc... which have been setup and defined in the cube.  This does not cache any data.

If anyone has a good resource or additional information, I'd love to hear about it.