Large process chain question

[Former Member]

Hi Experts,

I am trying to create a process chain that has around 25 infopackages, and i was wondering what else i needed to add to the chain apart from the "start", "delete index" and "create index". Also when loading master data in the chain do i need to add "change run" for each master data infopackage?

Thanks, point will be awarded.

[former_member188975 · Accepted solution]

Hi Gopan,

You can certainly string along the InfoPackages. Usually we have separate chains for master data and transactional data, with the master data running and activating first. This helps to make the transactional load a bit more efficient. Cases where master data load is included with the Transactional load is if there is some dependancy or look up in the code, and you want the master data to be fully updated and current.

If you are just doing delta loads one assumes that the voulme of records would not be too much, so in this case there is no need for the delete and recreate index process (as these would unnecessarily slow down the loading process).

You can add the change run in the master data chain at the end of all the loads (make sure you maintaint the proper variant for this process to include all the reqd master data objects). You will only out this process in the middle of the loads, if eg. master data load2 depends on the values of master data1.

Hope this helps...

[Former Member · Accepted solution]

Do i put this "change run" at the end of each Infopackage load, or do i just put it at the end of the chain and include all required master data objects?

[former_member188975 · Accepted solution]

Hi Gopan,

You can put it at the end of the master data load chain and include all the master data objects in it. But in case one master data load picks up values from another one, you need to activate the master data1, before loading the master data2, so that correct values are updated (this is not a common scenario).

Hope this helps...

[Former Member · Accepted solution]

Put th change run after all master data attributes and hierarchies (if any) are loaded. We have a chain for all master data attributes (ending with change run), and a separate chain for master data texts. For our transaction data, we have a separate chain for each application (Sales, Inventory, PP, etc.) and are controlled by meta chains. The meta chains control the flow of the execution of the individual application process chains.

Hope this helps.

[Former Member · Accepted solution]

Hey you don't have to put "Change run "for each infopackage ..

1.use "and" process at the end for the parallel process and have a common "attr change run " process there..and include required master data objects.

2.or .put them in sequence and put attr . change run at the end and include required master data objects...

..Regards,

Rakesh

Message was edited by:

Rakesh Kumar

[Former Member · Accepted solution]

For the text chain i am trying to configure it to have a 2 minute delay between every 4th infopackage load [there are sixty in total], how do i go about doing this? Do i have to put a "and" process between every 4th infopackage?

[Former Member · Accepted solution]

If you really need a delay, then you can right click on an infopackage process and select "Waiting Time.." (BW 3.5 and prior) or "Debug"? (BI 7.0+). Here you can put in number of seconds to delay.

Our text chain has close to 40 infopackages running in 4 parallel columns without delays and we have never had a problem.

Hope this helps.

[Former Member]

Hi Gopan,

You can follow the below steps.

1. Create local chain for all Master data attributes.

Develop failure event if required. It will notify which data load has failed.

2. Create local chain for text data ( one for hierarchies if required ).

3. Create local chain for all trasaction data load ( ODS + Cubes )

In addition to delete index, create index you can add 'activate ods data',

delete ods data, load failure events etc..

4. Create one meta chain which will start with 'start' event followed by master data local chain->text data local chain then add 'Attribute chain run' and after that you can add trascation data chain.

In above all chain you can use 'AND' , 'OR' operators as and when required.

Hope it helps.

Regards,

Aditya

[Former Member]

Hi ..gopan .

Its always a better idea to go with multiple local chain executed through a common metachain.

You can have classicfication ..which infopackages are for master data and which are for transaction data ,..so u can separate them into Master Data layer and Transaction data layer.

Also you can further bifurcate Transaction data layer in two layers..

1.Data load from source system to BW.

2.Data loads in BW to BW itself(like DSO to Cube).

This way you can monitor and resolve issues more effectively than keeping them in a big common chain.

Regards,

Rakesh Kumar