none
analysis services data extraction RRS feed

  • Question

  •  

    I'm working on a project and have tried numerous ways to solve this but nothing works yet.  Basic requirements: I need to extract all the data points from an analysis services cube... I have been successful at pulling all the metadata I need and I can get the data points out but it is slow.  Also, the data needs to be pulled with a set of time dimension cells in a specified order (the set has been created for simplicity)...

     

    I am using AdomdDataReader (and associated connection and command objects) and C# to get this done.. I've tried multi-threading it which sped things up and seemed great until i actually tried reading the data (the queries ran fast concurrently but the reader threw exceptions when reading).  I assume this is due to threading issues in the .NET framework from what I've read.  so, back to the drawing board...

     

    we are migrating our production environment from Oracle and Express to SQL Server 2005 and Analysis services... but express took 6 hours to process a moderate sized data set and MS is taking over a day to process the same data... I'm sure there are optimizations we can make to the cubes and I'm hoping there are better ways to access the data...

     

    any help would be greatly appreciated... or point me in the direction of someone who might be able to help... thanks in advance!

     

    Max

    Monday, December 3, 2007 8:04 PM

All replies

  • The cube is sources from a relational data source.  Do you have the option of pulling data from the source relational data warehouse?

     

    B.

     

    Tuesday, December 4, 2007 9:43 PM
    Answerer
  • i've thought of this approach but the cube handles all the aggregating and calculated measures etc. this would be fine if all i had to do was extract addative measures but it's more complex than that.  our old system used Oracle and Express (at least 10 years old) and it's discouraging to see that this new technology is so much slower than the old!

     

    thanks for the suggestion.

     

    Max

    Wednesday, December 5, 2007 4:52 PM