Dear all,
In our projects we've faced the following problem.
We have BPS user-exit planning function. This function should process a large amount of data. And, of course, performance of this processing is very critical!
To increase performance we divided whole data amount into subsets and launch mentioned PF for each subset in background within separated thread ([http://help.sap.com/erp2005_ehp_04/helpdata/EN/c4/3a7f1f505211d189550000e829fbbd/frameset.htm], [http://help.sap.com/erp2005_ehp_04/helpdata/EN/fa/096e92543b11d1898e0000e8322d00/frameset.htm])
But now the problem:
Processing time for first thread is more, than 15 minutes of 35.000 records.
For second one - 5 sec(!!!) for 25.000
Third thread - 4 sec for 20.000
and so on...
In some threads calculation time also increases to 10-15 minutes for one thread. Statistic of calculation is below
Thread - Ecex. Time - Total read - Changed - New
1 - 0:12:46 - 33124 - 16317 - 1252
2 - 0:00:03 - 12936 - 2202 - 264
3 - 0:00:05 - 17108 - 6072 - 468
4 - 0:00:03 - 8540 - 2184 - 402
5 - 0:00:03 - 0 - 0 - 0
6 - 0:00:03 - 448 - 56 - 180
7 - 0:00:04 - 8344 - 2250 - 390
8 - 0:00:04 - 14112 - 4184 - 404
9 - 0:00:05 - 24388 - 10190 - 560
10 - 0:00:03 - 11482 - 4728 - 606
11 - 0:00:04 - 11508 - 4740 - 616
12 - 0:00:04 - 0 - 0 - 0
13 - 0:00:03 - 5936 - 1168 - 180
14 - 0:00:03 - 15036 - 4900 - 662
15 - 0:00:04 - 17640 - 5516 - 366
16 - 0:07:18 - 18284 - 8810 - 572
17 - 0:07:05 - 17752 - 8064 - 436
18 - 0:00:00 - 19966 - 6018 - 384
19 - 0:00:00 - 0 - 0 - 0
20 - 0:00:01 - 19992 - 6016 - 384
21 - 0:00:01 - 336 - 82 - 180
22 - 0:00:00 - 0 - 0 - 0
23 - 0:00:00 - 11228 - 2302 - 444
24 - 0:00:03 - 224 - 64 - 180
25 - 0:00:03 - 7952 - 3510 - 180
26 - 0:00:03 - 8288 - 2058 - 390
27 - 0:02:30 - 11928 - 5292 - 180
28 - 0:00:04 - 19096 - 3964 - 486
29 - 0:00:00 - 1568 - 490 - 180
30 - 0:02:29 - 11452 - 2428 - 444
31- - 0:00:01 - 12656 - 3436 - 504
32 - 0:00:00 - 0 - 0 - 0
33 - 0:00:09 - 336 - 126 - 180
34 - 0:00:08 - 4648 - 1086 - 180
35 - 0:00:07 - 0 - 0 - 0
36 - 0:00:06 - 0 - 0 - 0
37 - 0:00:05 - 336 - 24 - 180
What is a reason of such behaviour of PF?
Where is a bottleneck and how to eliminate influence of bottleneck on execution time of whole data amount?
Your advices