9117-MMA Performance
Home > IT Answers > AS/400 > 9117-MMA Performance
Tonks
5 pts.
0
Q:
9117-MMA Performance
9117-MMA, OS/400, V5R4, DS6800
WE have recently installed a 9117 MMA p570 which has 11 partitions, 3 of which are iSeries. All disk is on DS6800. Our performance has gone backwards on the Production iSeries partition, runtimes blown out excessively though I have seen some batch jobs running faster. We  have heard thru the grapevine tripling the memory gets better results which we have tried and had mixed results. IBM are wary this may be just masking the problem throwing extra memory at it. Since turning QPFRADJ off we have had better results. Found where increasing the Max Active in the BASE Pool helps but to what .... 1000, 10000, 20000 ??

Any thoughts on this please Performance Tuning Guru's ??!!!



Software/Hardware used:
OS/400 (V5R4), DS6800
ASKED: Oct 13 2009  4:07 AM GMT
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
RELATED QUESTIONS
0
Whatis23
4040 pts.
0
A:
 RATE THIS ANSWER
0
Click to Vote:
  •   0
  •  0
  • AddThis Social Bookmark Button
It's surprising your performance improved after turning off QPFRADJ.
Before randomly changing the MAXACT in any pool, verify the average amount currently running in each pool.
Type WRKSBS and note which subsystems are running in each pool.
Press F11 to view the total number of active jobs in each subsystem.
Based on those totals, compare to the value for MAXACT in the WRKSYSSTS screen.
If any of those values are too low, some jobs will go to INEL state.
Increase the value to a number above the total you viewed in the WRKSBS screen.
If they're already above the total, that is not your issue and increasing the number won't help but be sure to use the average for the day(s).
You should start collection services to view daily reports of where the performance problems are occurring.
Last Answered: Oct 13 2009  10:33 PM GMT by Whatis23   4040 pts.
  •   
0
0
RSS - Discussions Help
Discuss This Answer:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _



_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Graybeard52   2425 pts.  |   Oct 14 2009  5:16PM GMT

Performance can be a tricky thing, because it all depends on so many factors.
My experience with Performance Adjuster was worse than yours. I am running several MQT builds during the night time hours over some large files with 60-70 join files. Suddenly, run time want from 2 hours to 24+ hours (did not let it finish). Here is what the problem was: SQL optimizes based on available memory, but does NOT lock that memory. Consequently my system was optimizing based on about 6G of memory. But other jobs running in QBATCH eventually consumed most of the memory, and SQL went to a near stop trying to run in justa little memory. Perf Adjust just kept pulling it away from SQL.
My solution was to run SQL thru a seperate memory pool, with 2G minimum and 2G maximum. Now Perf Adjust does nothing with it, SQL optimizes at 2G and gets to keep it for the entire run. Those queries are now down to about 20 minutes each, and smaller queries have benefited as well.

 
0