High volume BP Data Loads 1M+ to MDG and DRF to S/4 HANA 1909

[gopijags23]

Need some information on Business Partner Data Migration to MDG for very large volumes and the best practices around it.

We have a customer with 1M+ Business partners in legacy systems and we need to decide on the best approach to load data to MDG(Hub based) and S/4 HANA 1909.

1. If we load 1M + BP to MDG first and use DRF to replicate to S/4, how is the DRF performance for such large volumes into S/4. Customer wants to know how long has it taken for other customers, so it can help us extrapolate and predict load times for 1M records.
2. Is loading into MDG and S/4 separately and then reconciling the Keys and GUID via Key/value Mappings a better/preferred option, than DRF to S/4, for such high volume scenarios?

Any inputs will be very helpful on the above 2 questions.

Regards

Gopi Jagarlamudi

SAP America

[mageshwaran_subramanian]

Hi Gopi,

Here's my answer based on my experience. We loaded ~50K customer records into MDG hub (1809) and then replicated to ECC with DRFOUT. Please note that CVI was set up to generate the BP in the background.

1. 50K customer records (DEBMAS, ADRMAS, ADR3MAS) took about 3 hours to load into MDG - Data Services was used for the load. DRFOUT was pretty slow. Took about 8 hours to complete the replication. In your case you would be replicating the data into S/4 - this may be faster.

2. I would say YES. Load the data twice a. Initial data load into MDG (create) b. Load the same data into S/4 again (Create/Update based on the scenario)

Happy to answer if you have more question. Keep this thread posted on how the load goes.

Regards,

Magesh

[former_member230136]

Hi Magesh,

I was following this thread wondering what can be done best to achieve that. By looking at your answer, I am now curious how did you manage to pump 50k Customer Records to MDG using Data Services.

Point 1.

For us:- Customer Initial Load took almost 30 hours for pushing 130k Customer Records into MDG System( DEBMAS, ADRMAS & ADR3MAS) IDOCs were used to load the data. Our MDG System is also HUB running on SAP ERP.

Job was sending IDOCs in parallel processing.

Did you also use Parrallel Processing for sending IDOCs, what was the Batch Size which you had opted for ?

I am asking, as it is good to know if we can achieve optimization during the load.

Point 2.

If this is a greenfield implementation, why loading the data twice considered a good option. In my opinion, as this is S/4 and BP Replication is required. We should be using SOA for the replication keeping the keys similar in both the systems as this is a recommended approach. Again I am not sure how the performance will be here in case of 1M+ records.

Thanks,
Rohit

[mageshwaran_subramanian]

Hi Rohit,

On point 1, during the mock loads we identified that below two services causing the slowness in idoc processing. Deactivating these services for the initial load improved the performance. Also, 1. Sent the idocs from DS to MDG (64 status) 2. Processed the idocs in MDG. Parallel processing - yes, we didn't see packet size has any impact our load performance but it was set to 200.

You can refer to this doc to know what these functionalities do.

On point 2, I believe we can keep the keys similar in both system even by loading the data twice. I would suggest this only when the replication takes long time to complete (during the mock load and replication)

[former_member230136]

Thanks Magesh for the response.

Good to know the optimization which you did in order to speed up the data load.

Cheers,

Rohit