HELP!!!! - Serious OW Performance Issues

jdcmwilson

Well Known Member
JDE List,

I AM EXPERIENCING SERIOUS ENTERPRISE SERVER PERFORMANCE ISSUES. When trying
to do open sales orders loads into OneWorld using R47011 it takes me 1 hour
to load 600 records. A couple of weeks ago I was able to get 4000 loaded in
an hour. In comparing Oracle reads on the Dev box to Oracle reads on the
Prod box, the R47011 UBE does between 700 to 800 reads in the same amount of
time that the Prod box does between 15000 and 16000 reads. What do I look
for here? Is this Oracle, OneWorld, ???? We are not aware of what has
changed in the last couple of weeks to degrade our systems.

One thing that has brought this performance issue to light is that our
Development Enterprise Server runs much faster than our Production
Enterprise Server. Our Dev box is an IBM RS/6000 M80 with 2GB of RAM and 2
processors. Our Prod box is an IBM RS/6000 H80 with 8GB of RAM and 8
processors. It doesn’t take a rocket scientist to see that our Dev box
should not be performing markedly faster than our Prod box.

The only thing that I can think of that is different is that the PD7333
Central Objects on our Prod box were not created using the OneWorld scripts.
We attempted to do so and had problems so we created the COs manually.
All of our Control Tables and Business Data were copied from an environment
that already existed on our Dev box. Many times we do data copies from
environments on our Dev box to our Prod box using Oracle rather than JDE’s
R98403. It just seems to work faster and more reliably (is this part of our
problem?).

Additionally, we began with Security, System Tables, Data Dictionary and
Object Librarian on our Dev box but have since moved all but Security to our
Prod box. We are planning on making our Prod box the Security server
tomorrow. Should this make any difference?

I don’t know where to go next as we have been scratching our heads for some
time on this one. Any ideas, things to try, brain-storms, “gotchas” or any
other items to investigate would be greatly appreciated. We are planning
(trying) to go live on Oct. 1 but these performance issues are impeding our
progress.


James Wilson
CNC Consultant

OW Xe B73.3.3 SP16 AIX
IBM RS/6000 AIX 4.3.3 Enterprise Server (Central Objects)
Oracle DB 8.1.6.3
NT 4.0 Deployment Server
"fat" Win95/Win98 and TSE/Citrix "thin" clients
ZenWorks - using to deploy initial OneWorld clients



All views are expressly my own and in no way reflect those of my employer.
 
James,

this is obviously late but we (lister's) just saw your plea.
One suggestion is to check the PD7333 central object tables to ensure that all indices exist and are valid.

Otherwise if things work OK (just slower) you've something else boogered up in your configuration. Check and compare the following settings from your backups of a couple weeks ago:
- UNIX Kernel Parameter
- Oracle DB Parameters
- Oracle Listener Parameters (Oracle s/b using IPC locally not TCP/IP)
- OneWorld Server JDE.INI settings

Also check to see if table F986110 has a very high amount of reads.

My 2 cents,

Larry Jones
[email protected]
OneWorld XE, SP 15.1
HPUX 11, Oracle SE 8.1.6
Mfg, Distribution, Financials
 
James,

Well, if performance gains decreased on my prod box, meaning it was performing well before, then something had to change, get filled, etc...

1.) Check to see if you have logging turned on, on the Enterprise server. Debug? Ube? Sounds simple enough, but it happens.
2.) Row/table security turned on? Since your F00950 table is on a different db, there may be an IO bn.
3.) Are the sizings and properties of your db the same? How is your temp space? Logs?
4.) During run-time, your central objects will not play a role. Regarding your copy of the co db (or any db for that matter) through oracle and not using R98403, I haven't had any problems. Obviously, you have to update the F983051 and F9861 tables. If you used a bulk export/import by table owner, Oracle should have rebuilt your indices.
But, **Note** The issue of indices in Oracle on XE is also an issue. Turns out by default XE puts the pk in the data tablespace by default and not in the index tablespace. What you wind up with is separate pk's and indices, that invariably give you performance degredation as the data is grouped with the index. There is a good doc out there about this with some great scripts, let me know if you don't have it/need it.

Without much more info, I can't be more help, sorry!

Hope this helps,

-mo
 
Hi James,

Have you tried ANALYZE TABLE in oracle?

The statement eg:
ANALYZE table PRODDTA.F45715 COMPUTE statistics;

We have had a similar issue - one of our enquiry apps used to take 2 secs, then suddenly it took 5 mins. After the analyze it is back to 2 secs.

Regards,

Gopal.
OW XE U1 SP 14.2, Oracle 8.1.6.3.0, Sun Solaris 7
 
James :

Have you tried to reindex and update statistics on your PD databases?

Sebastian
 
You were a huge help. Would you be so kind as to provide me with the "good
doc" with the "great scripts"?

Feel free to send them to [email protected].

Thanks,

James Wilson
CNC Consultant

OW Xe B73.3.3 SP16 AIX
IBM RS/6000 AIX 4.3.3 Enterprise Server (Central Objects)
Oracle DB 8.1.6.3
NT 4.0 Deployment Server
"fat" Win95/Win98 and TSE/Citrix "thin" clients
ZenWorks - using to deploy initial OneWorld clients








All views are expressly my own and in no way reflect those of my employer.
 
Back
Top