How to reduce a cube size

[myshah789]

Hi techies, please guide me for the below questions:

1. If my Cube size is 2GB, now I have got the requirement to reduce it. What are the parameters I have to follow to reduce the cube size?

2. How to get the 2nd highest salary of an employee without using any function (I mean without rownum etc.)?

3. What is the difference between joins and scope relationship?

[cognostechie]

Interview questions ?  Not to worry..

1. Hold the Cube tightly in your palms and start squeezing it. The more you squeeze, the smaller the Cube becomes. That's how you reduce the size of the Cube.

2> You never try to get the 2nd highest salary because that employee is not worth anything. Only worry about the highest salary employee. Does anybody remember who was the second person who landed on the moon ? Everybody remembers the 1st one..

3> Join means like mjoining hands with somebody to form a business deal whereas scope relationship means the extent of the relationshiop (Ex: The scope of two people married to each other is the relationship of a husband and wife. As you know the scope of a girlfriend/boyfriend will be somewhat less).

Hope it helps ...  ;D

[MFGF]

Hi techies, please guide me for the below questions:

1. If my Cube size is 2GB, now I have got the requirement to reduce it. What are the parameters I have to follow to reduce the cube size?

2. How to get the 2nd highest salary of an employee without using any function (I mean without rownum etc.)?

3. What is the difference between joins and scope relationship?

1. What kind of cube? Powercube? TM1 cube? MSAS cube? They are all different technologies and are optimised in different ways...

2. Why is there a restriction not to use a function? I would suggest ranking the salaries using a rank() function and filtering for Rank = 2

3. Joins are relationships defined in a database or in a Framework Manager, specifying how tables/query subjects link together. In FM they also define the cardinality.  Scope relationships define how measures are in scope with the levels of a dimension from an OLAP perspective. In FM they are created and used between Regular Dimensions and Measure Dimensions.

Regards,

MF.

[myshah789]

1. What kind of cube? Powercube? TM1 cube? MSAS cube? They are all different technologies and are optimised in different ways...

2. Why is there a restriction not to use a function? I would suggest ranking the salaries using a rank() function and filtering for Rank = 2

3. Joins are relationships defined in a database or in a Framework Manager, specifying how tables/query subjects link together. In FM they also define the cardinality.  Scope relationships define how measures are in scope with the levels of a dimension from an OLAP perspective. In FM they are created and used between Regular Dimensions and Measure Dimensions.

Regards,

MF.

Hi MF,
First of all, thx for ur quick turn around.

I have a Power Cube which is very huge (approximately 2 GB), it's offering a huge performance. Now I'm asked to reduce the size so atleast I can improve the cube performance at some extent.
I'm thinking for incremental update and partitioning........
That's why I want to know that what are the parameters I have to follow.

Please advice!

--Shah

[absriram]

Hi Shah,

How many dimensions and levels do you have?
Check the levels with the maximum number of categories to see if all the categories are required.
Check your hierarchies to see that there are no duplicates. For example, if you have a slowly changing dimension then the same level might appear twice. Make sure this doesn't happen.
Last but not the least, ask the users, which levels are needed in the cube. The cube should be used for aggregate level information and you may be able to exclude levels that are very detailed and generate a lot of categories.

If you have checked all these and still your cube size exceeds 2GB, partition your cube based on the time dimension. This will break up the cubes into child cubes, for example, one per year, quarter, month or even a day, depending on how much data you have.

Hope this helps.

Sriram
http://cognosonsteroids.blogspot.com