E9.1 MRP Performance

[jhsiao520123]

Dears,
When client was in CRP phase that not much data (100 items and simple BP relations) so R3483 can be finished in 2-3 minutes. Now we dump all 2500 item + 500 relationships + 18 month weekly forecast + horizon 4W+35M. We run all items at one shot but now never finsihed (I terminated couple times try to revise the PO/DS to run again but no good)
Noticed F3411/F3412/F3413 keep increasing to the huge numbers (at least over 10M for F3412 and keep going). We have 4 BPs and relates to each other for RM and also FG. And have looked Oracle document for reference but no specific to be our issue.

The BOM are not deep (general 3 levels) but with all bulk level items set to be with batch size (Chemical) to the BOM. I don't know if any specific reasons the data related cause the issue, or some other techinal issue can be. I have clients are with much more data than this but MRP (R3482 however) can be done in 2 hours. Any your practice experiences can share? Thanks!

John
Apps9.1 Tools 9.2.3.1

 

[DSauve]

One thing you might check is to make sure you don't have any Item Master or Item Branch records with a short item number (alias ITM) of zero. We've had these appear once in a great while and they significantly impacted MRP performance.

 

[jhsiao520123]

Hi Lists,
Found the reason and finally the R3483 finished with 4 hours. It used to build the batch size BOM with "V" components and this makes system multipling the batch size level by level to explore and so got huge numbers of F3411/F3412/F3413. After changed the "V" to "F" we got all outputs of MRP.
The parameters are with horizon 4W+23M and forecasts of 8 weekly + another 16 monthly, and we are still trying to cut 4 hours further!
Regards,
John
Apps 9.1+TR 9.2.3.1

 

[jhsiao520123]

Here is interesting my first time run R3483 took around 4 hours however today ran 2nd and 3rd time (all seetings are the same) with more F3411/F3412/F3413 records wrote but tooks 7 and 9 minutes only. I ran all 3 times are full generations but not net change for 2nd and 3rd. Can anyone shed the light why the time needed can be so different? The table index need to be built at the 1st time run only? Confused!
Regards,
John
Apps 9.1+TR 9.2.3.1

 

[Larry_Jones]

No real explanation other than I was amazed that your small data set would take 4 hours to run. 7 - 10 minutes seems more likely.