Multi Dimensional Modelling in Teradata

[rohitrevo]

Most of the databases like SQL Server 7, Oracle 9i, IBM Essbase have a multidimensional engine which allow OLAP analysis. But Teradata V2R4 does not have a multidimensional engine and if we have to make cubes we resort to cube building in OLAP front end tool like Microstrategy and Cognos PowerPlay. But this approach has its limitations.

I think the absence of a multidimensional engine in Teradata is a big disadvantage. Are there any methods to work around this disadvantage?

 

[BillDHS]

Athough I am not an expert on cubes, there used to be a product called TeraCube being developed with the Teradata OLAP functions. I have also read articles that suggest the use of aggregate / summary queries and tables to use instead of cubes.

I am not personally a fan of cubes. They take up a lot of space for values that may never be aggregated, and they can result in inaccurate values by summing up inappropriately.
For example we want to get a number of distinct clients. If we report distinctly at the County level, than roll up distinctly to the State level, most cube would give an inaccurate count by just summing the County level and not exclude duplicates.

 

[tdatgod]

Hi,
We sold teracube lock stock and barrel ( including the 33 developers associated with it) to Micro Strategy in October of 1999.

What they did with it I don't know.

 

[tresgatos]

Microstrategy has continued to develop the product. They have strong ROLAP support for teradata. And thanks to the teracube software purchased from NCR you can now talk to microstategy using OLEDB/MDX.

 

[tresgatos]

And one more thing....
If you are using ODBC to talk to teradata, I don't see why you can't build cubes pretty easily using MS Analysis Services. I haven't tried it, but people have used it to work with Oracle, so I don't know why it should be any harder to use it with teradata. That would be your #1 cost-effective thing to try (unless you already have microstrategy)