BW Logical Partitioning

[former_member1213279]

BW Portioning

Hello Experts,

I would like to perform partioning of one our cube which is approximately having 100000000 records.

We are having data since 2005 till current date.

I have done partioning of cube from 01.2008 to 12.2009 , after doing this when I checked in background I found that there are 24 partitions has been created.

1ST Partitions : 01.2008 to 12.2009

How can I add new partitions to existing partitions?

When I tried to add new partition 06.2010 to existing one (01.2008 to 12.2009) it has overwritten existing one.

My query is if we create large number of Partitions then does it going to affect data load timing?

If some one knows proper way to do partioning then please suggest, my data volume is 100000000, range 01.2005 till current date

Regards,

Pankaj Naik

[former_member1213279]

Hello Experts,

Thanks for your Help!!

Cube is showing that my it is already partioned Based on 0CALMONTH for 012003 to 092009.

We have already exceeded the validity date for range partitioning and for that a single, new, and ever-growing partition will be created, which will contain all the latest data and all the previous data since the validity date was exceeded. i think this can lead to performance problems since the partition can grow quite large and will be accessed continually for the most recent information. It will contain all of the data added to the cube since the validity date was exceeded.

So when i tried to add new partions it is not allowing me to do this.

1. Can you please help me out for repartioning of cube.

Also i am having one more doubt.

This cube conatins data from 2003 so if we create new cube and distribute data from exting cube to new cubes based on CALMONTH lets say for 2003 to 2005 A1 cube, 2005 to 2007 A2.

To do this we can move data from Exting cube to A1 and A2 based on CALMONTH and then perform Selective Deletion on Actual Cube.

And then creation of Multiprovider baseod on Existing cube + A1 Cube + A2 cube.

Can you please suggest me which one will be good option

1.Partioning

2 Craetion of Multiprovider on 3 Cubes?

3 Both option 1 and 2

Regards,

Pankaj Naik

[Former Member]

In your case, the cube is holding far too much data.

Your data goes to max. partition as it has run out of its limits.

I do not think its possible to add a new partition, unless you drop data. (I mean physical partition.)

It would be advisable to create a couple of new cubes and either transfer the old data into that cube or keep the data from 2003 to say 2005/ 06 in this cube & transfer the new data into the new cubes.

Once the cubes are in place, create a Multiprovider on these.

Also, partition the cubes based on 0CALMONTH & if possible, have 0CALMONTH as filter in your queries.

Create a repartition process chain - this is repartition the Aggregates in the cube.

Useful when you have exceeded your 0CALMONTH limit & you do not require data, older than some fixed date, for reporting.

The data will remain in the cube, but it can be deleted off from the Aggregates.

Alternatively, if its possible, you can also go for SAP BW 7.3.

There are some options for partitioning even if the cube has data. I am not sure of these, but surely you have better options.

[Former Member]

Hi,

why dnt u create another cube and stpre the data. In one cube you can store only 3yrs of data while the other cube can hold the historical data. Then you can report on a miltiprovider.

Regards,

Rahul

[Former Member]

Is it necessary that you have data in the cubes from 2005?

You can do the following:

1. Create a Copy cube of the existing one and load the Historical data into that cube (say, from 2005 to 2008).

Create partition on the cube depending on 0CALMONTH.

2. Aggregate Repartition for your existing cube - create a process chain with the following steps:

In this case, you can set the time line from which you want the data to be partitioned. Data falling outside the timeline will be deleted from the aggregates.

Start -> Selective Delete -> Merge Partition -> Add Partition -> DB Stats (to check the data in the partitions on D/b)

In your Selective Delete step, mention the number of periods in History & Future which you require the partition on.

Execute the Process chain & your aggregates will be re-partitioned.

3. Cube Partition - For this, you'll have to delete the data from the cube & set the partitions again.

I do not think this fits your requirement.

I hope this helps.