MRP Job taking more time

Question

Dear Folks,
We run the MRP at MRP area level with total 72 plants..Normally this job takes 3 to 4 hour to complete...but suddenly since last two weeks it is taking more than 9 hours. with this delaying business is getting problem to send the scheudles at proper time and becomes critical issue.
Now anybody have idea to check the root causes of this delay? And how to reduce the time... We have already processing this job with parallel processing.
Reasonable answer will get full points.
Regards
TAJUDDIN

Answer 1

Hi TAJUDDIN
Unfortunately, I do not have any documents related to parallel processing.
But I can explain how to do, So I hope following explanation help you.
1. At first you need to check whether current parallel MRP works fine or not.
    To do this, please check MRP result in the spool.
     (In the last page of spool, you can see the result task use of each WP.
      To open last page of spool, just go to SM37 and find MRP job and then
      press spool button. Now you will see spool overview.  But before open
      spool please change page setting that you will see.
      <From the menu,  Goto=>Display request=>Setting> Then select last
      10 page.  By doing this operation, you can see last 10 page.
2.In the button of the spool page, you will see task use of MRP.
   For example. if you use 2 application servers and each application server,
   if you assign 3 work process, you will see 6 task.
                         Number of calculated item   Runtime
   AP1 WP1        1000                                  30 min
   AP1 WP2        500                                     5  min
   AP1 WP3        200                                     3  min
   AP2 WP1        200                                     3  min
   AP2 WP2        200                                     3  min
   AP2 WP3        100                                     2  min
  In the log, if you observed above situation, this indicate unbalanced system use.
  (This situation occur depending on your BOM structure). So there is a possibility
  that dispatching system use more equally improve MRP performance.
  To do equal dispatch, you need to de-activate MRP package logic to  bottlneck
   Lowlevel code.(You can see bottlneck item in the spool.  So if you observe
   10 items belongs to lowlevel code 5, it is better to deactivate package logic
   for lowlevel code 5. Then for this lowlevel code, no package logic works and this
   bring to equal distribution of task use).
The way to de-activate is described in SAP Note 568593  and 52033
   (Until 46C, you need to use modification <manual change coding>. From
    Enterprise, you can use BADI).
Regarding on package logic, I recommned you to read SAP Note 52033
(Depending on runtime of former task, MRP combine several item into 1 package.
So if task 1 finish previous MRP around 60 sec, next MRP for this task will contain
  more material <package comes to contain more material>. But if bottlneck items
  are put in togather in 1 package, this task 1 might take more longer time
  cpmpared to others.  And MRP calculation occur low level code by low level
  code. So if this task 1 does not finish calculation < other task already finish
  MRP calculation>, other task cannot start MRP for next lowlevel. So other task
  have to wait until task 1 finish his MRP.  Due to this, you will see big task usage
  and runtime difference in spool).
But this behavior depend on BOM structure. So you also have possibility not to see
this behavior in your spool. In this case, no necessity to consider balancing..
I hope this help you.
best regards
Keiji