Cube phisical representation 1

[yoanka]

Hi,
Can anybody answer me this question:

How is the .mdc file built? Does it have a certain implemetation inside? I'm asking because of effectiveness of large cubes.

Let me give an example: If we have a cube with many dimensions and not many values, so lets say out of 10 shops each one sells only in one region and eight of them have sales only in last two years - in many intersects we have missing values. Do we have in the .mdc file a record representation rather than multidimensional table (that would to be more effective in that case) or the way of saving data in .mdc is totally different?

If you have any idea about it, although I know this is Cognos know-how that is not revealed to users...

Thanks!
Greetz, Joanna

 

[DoubleD]

I'm not sure exactly what you're asking for in this post, but hopefully this will be somewhat helpful.

If you have many blank values I would ask the following questions:
1. Is Powerplay the right tool?
2. Do I have the right dimensions in this cube?
3. Should I have several cubes that answer the different questions I have?

You may be better off using prompted Impromptu Web Reports to represent your data.



Pain is stress leaving the body.

DoubleD

 

[flex13]

OLAP tools will build a series of intesections in their cubes. For example, if you had the following:
- Company ABC
- Canada
- 2003 revenue = 12,000
- 2004 revenue = 23,000
- 2003 cost = 10,000
- 2004 cost = 20,000
- Cars

The physical representation would be something like:
Company_ABC.Canada.2003.Revenue.Cars=12000
Company_ABC.Canada.2004.Revenue.Cars=23000
Company_ABC.Canada.2003.cost.Cars=10000
Company_ABC.Canada.2004.cost.Cars=20000

You can see that for every new measure a whole new series of intersections must be built. This is what really increases the size of the cube. If you have 100K categories, with one measure, this would create 400K intersections (example, not actual number). If you added one more category, this may only create 4 more intersections. But if you added one more measure, it would create another set of 400K intersections.

As you can see, the amount of measures really affects the size of the cube thus impacting performance as the amoutn of information within increases.

Hope my babbling helps ...

 

[yoanka]

Thanks flex13. That is exactly what I needed to know.
Greetz,
Joanna