Runbatch using excessive CPU

Hi List -

When the Runbatch process run on my enterprise server it consumes between 20 - 50% of the processor. I several multi-threaded UBE queues and when multiple UBEs are submitted the CPU gets pegged at 100% causing interactive users problems.

Any suggestions on how to lower the amount of CPU utilized by the Runbatch processes would be appreciated.

Thanks,
Bill

How many CPUs and how many *total* UBE threads do you have?

My enterprise server has 2 processors and this is the way the UBE queues are setup in my JDE.INI.

UBEQueues=10
UBEQueue1=QB7333
UBEQueue2=QB7333
UBEQueue3=QB7333
UBEQueue4=QB7333
UBEQueue5=QBATCH
UBEQueue6=QBATCH
UBEQueue7=CHARTER1
UBEQueue8=CHARTER
UBEQueue9=EOMBATCH
UBEQueue10=QB7332

Bill :

Check your SQL indexes. I've noticed that poorly indexed tables
cause excessive CPU usage by running UBEs.

Sebastian Sajaroff (JDELIST).

Bill,
have you checked if the debug is not turned on on your ES?
What is the value for QueueDelay in your server´s JDE.INI (the default is
30)?
How many lines are there in svm7333.f986110 (Job Master Table)? Are the
indexes to this table correctly build?

Even with several queues running idle the CPU consumption should not be
very high at all, close to 0% .

Thanks, Gerd

And normally, how many threads are concurrently processing?

Normally there are 2 threads doing concurrent processing.

Bill

20% for a single UBE is not unusual. I've seen financial reports take 90-100% of a single CPU for hours at a time.

Two simple options. (There are many more.) You can,

- Reduce the number of queues so that reports do not monoploize the CPU. In my experience this has not made users happy.

- Install an application server and move either the UBEs or the mapped BSFNS to this new server.

Increasing the size of your server would help, but, I've found that as long as you run both UBEs and BSFNs on the same server UBEs will always degrade the speed of your intereactive users.

Have fun.

How many RAM do you have ? We had the same kind of issue... We add more RAM and fix the problem. The problem is if your server is "swapping" a lot, your server will slow down. OUr server had 1.5 GB of RAM and 4 CPU's, we upgrade the RAM up to 5.5GB... It's working fine now !

Maybe a solution...

There is no way that I am aware of of restricting any process to a particular
amount of cpu time on NT/win2k.
Obvious solution, more cpu's, or less concurrent processes in the queues. Or
even better, get a logic server and offload the ube's to this.
Regards
Kieran Fitzgerald

No more than two UBE’s per NT CPU should be run concurrently for maximum throughput.

Any concurrent usage over two per CPU slows down all of the UBE’s.

If you are running UBEs, BSFNs and the RDBMS on the same box I think it is time you investigated a $5,000 Intel server to handle UBE load. I firmly believe that this is the single best performace increase per dollar for a OneWorld installation. Seperating the UBE load from everything else is beneficial.