Implementing APO DP Module

[Former Member]

We are implementing Demand Planning where the Planning team will be entering the consesus forecast through <b>SDP94</b> Transaction.

Currently its around <b>300</b> with a likelihood upto <b>500-600</b> within <b>1 year</b>.

We are planning for about <b>1200 WDs</b>.

We are doing the planning for around <b>1200 custome</b>rs and <b>800 SKUs</b>.

We are currently having <b>1 Planning area</b> and 1<b> MPOS</b>.

We have arround <b>250000 CVC's.</b>

We are facing high response time issues where the planners are doing a drilldown through the lowest level of <b>WD-SKU</b> combination.

Kindly let us know having multiple planning areas with multiple MPOS will improve on the response time.

[Former Member]

the number of Planning areas would depend on what your business process and the interaction and ability or possibility (in the business scenario) to separate the data.

But i suppose this will not be the case

There are a few other ways of tackling performance

Review configuration of the Planning Book

-reduce the number of KF

- use specialized planning book to ensure only a few KF are loaded. just the basic necessary.

- Limit the number of periods. there is no need to display 24 months in history or future if you are interested only in 6 months

Be smart about the way Data is displayed

- Show history only if it is necessary

- keep KF in display mode only if necessary. changing to display mode from change mode takes a bit of performance

- Restrict selection to a maximum to keep the data set loaded as small as possible. If possible replace your drill down with special selection.

macros

- make sure the macros are not set as start or default unless required to be.

- create specific planning books for mass processing jobs

There is another way to make performance better... that will be by preloading data - that will be to create POS aggregates and match the characteristics in these to the characteristics you load and the level at which you load it. This helps because data is already saved in the aggregate levels and as long as you load in this level you will load quickly (saving will take a bit of time)

[Former Member]

Kaushik are you also having performance issues in loading and extracting data in and out of APO DP? Do you use SAP BW?

[Former Member]

Hi,

You can have one MPOS, one PLanning Area but different

Planning Books

.
Also try to restrict the data in the SELECTION IDS.
Create as many

selection IDS

as possible restricting the data to that particular user.

As I believe one user will not have the necessity to access all of the data.

Here are a few other suggestions which will increase the performance

(1) Planning Area:

Ø Key Figures:

If it is used for reporting purposes, move to BW

If it is used to calculate dynamically, consider planning book key figure

Limit the number of keyfigures in Live Cache as they have an effect on the performance.

(2) Planning Books design: Most Important- This is one of the reasons, the system interactive performance becomes slow

Reading Data from Live Cache:

Say a Planning Book has 10 Key Figures and the data view has 5 Key Figures.

Whenever we try to access the data view, it will try to read all the key figures from the Live Cache. This affects the performance heavily.

********

Ø It is always better to create as many planning books as possible without any issues.

Ø Hide Key Figures that are used for calculation

Ø Remove Unnecessary Horizons as there is a way of initializing the planning area for a single key figure

(3) Macro design: Audit the macro definitions

Ø Only calculate the horizon which is necessary

Ø Row Attributes should have only one iteration

Ø Default Macros: Consolidation of steps

(4) Batch Processing: Another important factor for poor system performance

Ø Create separate planning books for the batch jobs

Ø Combine macros which use similar key figures into a single job

(5) Sizing of the LC server: Factors which affect the sizing are key figures, characteristic combinations, horizon. Generally it is recommended that the memory consumption should not be greater than 80% of the available server memory space.

Hope this helps

Regards

Kumar Ayyagari

Message was edited by:

Kumar P Ayyagari