Slow OneWorld Xe: how to troubleshoot it?

We have a client with OneWorld Xe, Update 1, SP16.1, Windows 2000 on DS and ES, Oracle.

All client machines on the system get slow response on a random basis-- by "slow response", I mean that sometimes even changing between menus takes a minute, and opening applications goes slow as well-- even for applications that they are not JITI'ing. Everyone is running full packages.

On the other hand, sometimes the system runs just fine.

I'm looking for some ideas for how to troubleshoot this.

Up to now, I've tried the following:

1) Ensured that Debug=NONE in JDE.INI on the ES, DS, and client. Good.
2) Checked that DS is turned on and can be pinged from the client. Good.
3) Checked that the client JDE.INI has the ServerHelpPath set correctly and that the path can be reached from the client-- also we commented out ServerHelpPath on the client. No change.
4) We checked Minimum Technical Requirements. They're good on Service packs, memory, CPU, etc. **HOWEVER**, they have only 105MB free on the C: drive, where Win2000 is running. They have other disks, and have assigned virtual memory on other disks as well as on C:
5) Oracle client is the same as server.
6) Network seems fine.
7) They do have two descrepancies in OneWorld-- they do not have Media Object Queues set correctly (they all point to SERVER still), and they're Object Manamement Configuration still has User Overrides pointing to Control Tables, which is wrong (according to JDE, they should point to Central Objects).
8) We checked their backup schedule and asked if there were any heavy processes going on during the work day. Backups take place after working hours and there seem to be no heavy processes during the day.

Can anyone give me some insight as to where to look next? Thanks in advance.

Tim :

Check that antiviruses are not checking the following files :

*.xdb
*.ddb
*.c
*.h
*.lib
*.obj
*.bmp
*.pak
*.gif
*.ico
*.pak

There are literally thousands of these files on clients and servers, they're
not executable and antivirus activity imposes a burden on performance
because these are heavily used.

Sebastian

Tim,

can you provide more info on:

1) ES specs (CPUs, memory, network, etc)
2) Oracle configuration:
- Number of database instances (1 or ?)
- Parameters such as # of Buffers, etc
3) How many Batch Queues are setup?

Regards,

Try going to P98MOQUE on the gh9016 menu...this lists your queues for media objects, you need to make sure the paths are valid.

Thanks Larry, here's what the client has:

> 1) ES specs (CPUs, memory, network, etc)
CPU - Genuine Intel, 702MHz
RAM - 2GB
Network - 100Mbps
Disks -
JDE Oracle Tablespaces/Indices in one disk w/1.9GB free
Windows on C:, 2GB w/ 105MB free (that's worrisome, though they do have virtual memory configured on other disks).

> 2) Oracle configuration:
> - Number of database instances (1 or ?)
One oracle instance
> - Parameters such as # of Buffers, etc
db_blockbuffers = 90000

3) How many Batch Queues are setup?
# batch queues = 6

Thanks in advance.

Thanks labluez.

We noticed right off that Media Object Queues were not set. Would this affect performance radically? Unfortunately we cannot make this change right now (the client wants us to make recommendations, review the recommendations, form a committee, etc etc: this is Spain after all). But we will recommend the change-- we'd just like some asssurance that this might have a big positive effect.

Thanks very much.

For performance you may also want to consider;
1. splitting your application from the database server, increasing
your oracle sga and temp space for large sorts.
2. Split your sys, obj, dd, and svm from your production database and
put it into its own. This has given us tremendous interactive
performance and sign on times. Even if you cache a lot of your sys
tables in oracle, they wont stay there if you make changes to them and
many are used at sign on; data source, ocm and security.
3. Determine if your enterprise server is pegged on cpu and memory,
consider not mapping bsfns to the enterprise server until it is
properly sized.
4. Ensure that batch jobs are not using poor data selection like not
equal clause, this can cause full table scans on large tables.

Tim,

How is your network configured. I have seen several installs that had
frequent slow downs and it turned out to be that the Enterprise Server was
connected to a bad port on the switch. Also, you can see this if you are
running on a shared network instead of a switched one.

Andy


_________________________________________________________________
Help STOP SPAM: Try the new MSN 8 and get 2 months FREE*
http://join.msn.com/?page=features/junkmail



CNC Consultant
AS/400, SQL 7/2000, TSE and Citrix

OK Tim, more questions:

How many users?

How many CPUs? Sounds like only 1.

What is the DB_BLOCK_SIZE?

Assuming Block Size = 8K then 90,000 * 8192 = 720MB
Your Total SGA is probably between 750 and 800MB.
Thats a very high % of 2GB that is also being consumed by the OS + OW + all the user processes. Have you checked Paging?

Is the non-system disk a RAID drive?

Disk - where do the Temp files exist? On the System Disk (C?

Possible problems:
1. Only 1 CPU would be a major problem.
2. Insufficient RAM. Try cutting the Oracle buffers in half as a quick fix.
3. Low disk space - strange things start occurring when disk usage > 9x%

Net, net sounds like a minimum configuration - if so then the users are getting what they paid for.

hi timallen,
Are you running realtime virus protection ? you may want to turn this off on a workstation or two and see if this is causing your problem.
Regards,
Dave

Hi Larry,

> How many CPUs? Sounds like only 1.
They have two CPUs.

> What is the DB_BLOCK_SIZE?
8K

> Is the non-system disk a RAID drive?
They're on RAID 5 on the nonsystem disk

> Disk - where do the Temp files exist? On the System Disk (C?
Temp files are on C:

One more (IMPORTANT) issue: They are on Oracel 8.1.7, and they have the data sources configured with JDBOCI80.DLL. They should be on JDBOCI81.DLL, correct? Would this cause the massive slow-downs that they are experiencing?

Thanks in advance, and for all the help you've given us so far.

You didn't tell us how many users on this configuration Tim, could you?

We run Oracle on a HP UNIX platform - so its not the same environment as yours. We have tried both JDBOCI81 and JDBOCI80 with Oracle 8.1.6 and saw no performance differences. In fact our administrator has them set to JDBOCI80 currently.

I would worry about temp space on the C: drive, but I'm not real familiar with OW on a NT server. Someone else want to jump in there?

Turn on Performance Monitor (A Windows NT utility) and watch the server when the slow downs happen.

Good Luck!

> You didn't tell us how many users on this configuration Tim, could you?
Sorry, about 20

>We run Oracle on a HP UNIX platform - so its not the same environment as yours. We have tried both JDBOCI81 and JDBOCI80 with Oracle 8.1.6 and saw no performance differences. In fact our administrator has them set to JDBOCI80 currently.
Okay-- in fact another consultant just told me that this is not really a problem.

In the end, we ended up setting Oracle to USE_DEDICATED_SERVER=ON
in the sqlnet.ora. This seemed to clear up a lot of delays.

> I would worry about temp space on the C: drive, but I'm not real familiar with OW on a NT server. Someone else want to jump in there?
I, too am worried about this. The client will be changing the C: drive out to have more space, and to run on NTFS (it was on FAT-- is this a problem, anyone?)

> Turn on Performance Monitor (A Windows NT utility) and watch the server when the slow downs happen.
We did this, but unfortunately the problem is so sporadic (kind of a heisen-bug) that we couldn't make it happen as long as the performance monitor was on.

Thanks a lot for the help Larry.

Check out www.precise.com They have performance monitoring software for network and specifically SQL software.

You might be interested in this:
http://www.microsoft.com/usa/webcasts/upcoming/1472.asp.