Partitioning a Cube (contains data)

[former_member182470]

Hi,

I have a Cube which consists of around 4 Crore records(2007 to till date). I want to partition my Cube Yearwise to improve our Query performance. I am aware that we can split a Cube only when it doesn't contains data. But my requirement is different.

What would be the best way to Partition my InfoCube?

Could you please give me sthe steps to achieve this? It looks interesting...........

<removed by moderaor>

Regards,

Suman

Edited by: Siegfried Szameitat on Jun 7, 2011 8:40 AM

[waqar_ahmed · Accepted solution]

Hi,

You can creat multiple mini cubes year wise, like for 2007 you can create cube the only load 2007 data. Just create link between your main cube and mini cube and load the data using dtp. This will improve your query performance.

Regards,

ray

[Former Member · Accepted solution]

Hi,

well I would say do not create multiple 'minicubes'. Wait for 7.3 and the SPOs (semantic partitioned objects).

Siggi

[koleti_madhusudan · Accepted solution]

Hi,

You can do partition to improve your loading and report performance by following below steps

Steps:

1. First create new cube with same structure of your existing cube(which contains data from 2007 to till date)

2. Copy data into new cube ..then delete data from the existing cube...now existing cube is free

3. You can partition existing cube with 0FISCYEAR...existing cube will be partitioned year wise

4. Now copy data from your new cube to existing cube ( selection should year wise)

this way you can do partition

Hope it helps.

Thanks

Koleti Madhu

[former_member182470 · Accepted solution]

Hi Madhu,

I carried out the Creation of a copy of my existing Cube and loaded data to my new copy cube. Before partitioning the original cube, I need some clarifications on the following :----

---

1. In the course of creation of Copy of my Cube, I created Transformation and DTP. After this, I loaded data from Original Cube to Copy cube, but there was no Infopackage to run. I just ran Full load in DTP which got the entire data from original to Copy cube. Is it correct?

2. After I partition my original cube and try to load yearwise, How should I put selection year range without Infopackage between Original to Copy?

Please clarify me for the above........

Regards,

Suman

[former_member182470 · Accepted solution]

Hi,

If My Cube is already included in Multiprovider and queries, Will there be any impact in doing partition of this cube?

Regards,

Suman

[koleti_madhusudan · Accepted solution]

Hi,

1.in BW 3.5 , once you do export data source, it will create info package ...just trigger the info package

2. Partition doesn't have any impact on the reports (except performance), because you are not changing structure of cube

3. In BI 7.0 ,we can do cube partition even if it has data ( new feature)

Thanks

Madhu Koleti

[Former Member · Accepted solution]

Hi,

There will be no impact on Multiprovider and the Queries that are build on that cube.....Because if you run for perticular year then earliar they go and search entire data in the cube and fetch the particular year data...now after partition thay will directly fetch the particular year data from concerned partion which will improve the performance.....

Thanks

[gadhatharanthirunavukkara · Accepted solution]

Load the data from original cube to mini cube without any filter using full infopackage.After paratitioning the original cube load mini to original back without filters.Make sure that transformation is 1:1 Mapping.In BI 7 if you use adding partition or overflow partition it will paratition the incoming records not the historical record.

Thanks

TG

[former_member182470]

Thanks Guys

[Former Member]

This message was moderated.

[Former Member]

Can you do Aggregate Partitioning or Re-partitioning? Can be achieved by building a process chain.

But, have you defined partition settings while defining the cube structure? (Physical Partition?)

If not, then i am not sure how can you go about this.

Maybe you can create copy cubes, year-wise, and then load the data year-wise.

Then create a multiprovider to access the data through the query.