Performance problems --> High DB time

[Former Member]

Hello,

We are running ECC 6.0 on I5 OS v5r4m0.

We are having performance problems though our server has 5 CPUs and 16 GB memory. ST03 shows processing time much larger (about 3 times) than CPU time and DB time 80% of (processing time - wait time).

Our buffers are large so that almost no swap occurs (except program buffer)

What could be the problem?

Best Regards

Sezai Yigiter

[Former Member]

More basic HW information would be useful, please.

- What is the model of the iSeries in which you are running ECC6 ?

- What is the feature code of the disk controller card used (4778, 2757, 2780, 5580, etc.)?

- What is the feature code of the disk units used (4326, 4327, 4328, etc)?

- How many disk controlller cards and disk units are used ?

- Roughly, how many users are there using ECC6 ?

If you know Collection Service - Graph History tool in iSeries Navigator, please display the utilization of:

- Disk Arm Utilization (maximum) - guideline for good performance is 40%

- User Pool Faults (maximum) - guideline depend on the model and disk controller used

Satid S.

IBM Thailand

[Former Member]

Hi Sezai,

first of all would be interesting, what type of CPUs you are using. i5 HW with 5 CPUs and 16GB RAM would be totally misconfigured ...

Then you should check & apply note 428855. Then you should apply the latest PTFs as described in the infoapar (http://www.4soi.de/IBM-Infoapars.php) and delete all SQL Packs. Then you should turn off the db mon as follows and restart SAP:

as4/dbmon/enable = 0

When all of this was done, you should check for the times in ST03 and then should check in parallel in ST06 on the page faults in all pools ...

How high os the CPU utilization ?

I would say, more would be a special & individual performance check, that could be done shortly (after you did all the stuff above)

Regards

Volker Gueldenpfennig, consolut.gmbh

http://www.consolut.de - http://www.4soi.de - http://www.easymarketplace.de

[Former Member]

Hi Volker,

thanks for fast response.

I too think that there is misconfiguration on hardware but IBM does not confirm it. CPU utilization is rarely above 50%. Sometimes DB faults reach 200. As I know this should be 20/second

I will disable as4/dbmon/enable but should this job not work in production systems.

Hardware conf. of this system is better than planned PROD server

Also as you are an AS400 expert I want to ask you something.

How can I see CPU wait time?

wrksysact does not show it

Regards

[Former Member]

Hi Sezai,

I think, I could say, that I'm an AS/400 expert )

What type of CPUs do you use ? 830 ? 550 ? ???

faults: 20 per second is just rubbish

200 sounds for a 5way reasonable, but it all depends on the environment ... we will have to investigate more in detail.

it is very important to have less than 1 fault in Pool 1!!!

db-Mon: yes, it should, but on V5R4 it doesn't work at all, it only consumes ressources. IBM will fix this problem soon. Until then you sohuld turn it off.

CPU Wait time etc. is very special and I cannot explain this here via mail, sorry.

Regards

Volker Gueldenpfennig, consolut.gmbh

http://www.consolut.de - http://www.4soi.de - http://www.easymarketplace.de

Message was edited by: Volker Gueldenpfennig

[HuseyinBilgen]

Hi Volker,

We're planning to implement our ECC60 on iSeries too and wonder what kind of misconfiguration took place with Sezai's configuration?

What must be the optimum configuration as i learned from him that sometimes for different SAP actions (activation, transaction start, SE16 start) their system stucks and freezes.

At that moment they cannot even reach the AS400 system. The User Fault avarage at freeze time is over 7000 for pool 2 (sap). Is this normal, or what may cause this?

regards

[Former Member]

Hi,

I would say, that I do not have any idea, but by now was always able to find and fix these issues. But, this cannot done by such a group. This will be indivual consulting.

Just contact me and we will be able to fix that in 1 or max. 2 days consulting.

Regards

Volker Gueldenpfennig, consolut.gmbh

http://www.consolut.de - http://www.4soi.de - http://www.easymarketplace.de

[Former Member]

We saw this the first time we used ECC 5.0 - back in March 2005 (that seems so long ago!!). It appeared to be something to do with the Optimiser for DB2/400. We're not quite sure exactly what was the real issue, other than by creating an index (any will do) it seemed to get it back into action and the problems went away!!

[HuseyinBilgen]

Hi Alan,

Who solved the problem?

IBM or SAP?

But when they checked it, it looks the cause of freeze is random. Sometimes when they're activating a program, even I've tested and see that when i called SE16.

What what ot index or how to configure optimizer for DB4?

regards and thanx for your attention

[Former Member]

What you describe is exactly the issue we had, there was no logic to it, other than the long delay at getting things from the database. We had both IBM and SAP involved and we decided to apply an index to table MDUR which is the material reservations table which seemed to be the one getting the hit. It then went away - but we weren't totally convinced it was this - more the optimiser.

[Former Member]

Hi All,

Is dbmon now working properly on V5R3? We have had it turned off for over a year now and I feel like I'm working in the dark.

Thanks,

Craig

[Former Member]

Hi Craig,

I would say:

nearly ...

... at least as good in order to use it ... You were on my agenda anyway to call to let you know as the fix came out last week and everything except the returned rows seems to be correct.

the ptf you need: new version of si24467 (perhaps still test)

Regards

Volker Gueldenpfennig, consolut.gmbh

http://www.consolut.de - http://www.4soi.de - http://www.easymarketplace.de

[Former Member]

User pool fault rate of 7,000 should generally be considered too high if it lasts too long, say more than 5 minutes. What is the pool no.2 size and what is its associated MAXACT value (use WRKSYSSTS to see these) ?

An example of misconfiguration can be like :

- Using low-performance disk controller rather than the high-performance ones

- Using too few disk controllers, say only one rather than 2 or 3 depending on expected workload

By the way, you should make sure all patches of DW, LIB_DBSL, R3TRANS, TP, SAPOSCOL, etc. should be as high as they are available (say, higher than level 80 or so. I figure they may be at around level 100 by now. Assuming ECC6 on Kernel 7.00)

Satid S.

IBM Thailand

[Former Member]

>>> Sometimes DB faults reach 200 <<<

You should focus on total DB+Non-DB Fault rate. A practical value of less than 700 faults per CPU of POWER5 type should be considered OK in my experience (assuming high-performance disk controller- 2757 or 2780 or 5580 -is used).

Satid S.

IBM Thailand