cancel
Showing results for 
Search instead for 
Did you mean: 

Significant report running speed variance on different servers

Former Member
0 Kudos

We have two servers which are DBCI and DIALOG for our R/3 4.70 system.

We found that one of our query need to run over 9 minutes on DBCI server while it only takes 9 seconds on Dialog server.

The two servers are of exactly the same hardware configuration.

Anybody have any clue about this?

thanks,

Accepted Solutions (1)

Accepted Solutions (1)

Former Member
0 Kudos

Does the behaviour reproduce with every try?

I have seen the opposite lately, so in my case the dialog instance was a lot slower, than the DBCI. In our case the cause where thousands of sql statements generated be the transaction. The execution from the dialog instance took slightly longer per single statement, which added up remarkably in total.

In your case i suggest you check the cpu usage during execution and monitor os paging (swapping). Your db server might suffer memory shortage. A typical case would be, if that transaction uses a lot of memory on the SAP instance.

Other causes could be different parameter settings of the instances etc

Cheers Michael

Former Member
0 Kudos

Thanks Michael.

Yes, this behaves on every try. I've checked the parameter, CPU/swap usage and everything looks fine. We tested on zero system load coming out with the same result.

Former Member
0 Kudos

Which os do you have?

Can you give us more details about the query (transaction/report, SQL statement etc)?

Can you tell which processes are running on the cpu while you execute the query (either look in the task manager or run a top)?

Otherwise we will keep on guessing

Cheers Michael

Former Member
0 Kudos

OS:HP-UX PA-RISC 11.23

We have just tested all our reports and sorted out a list of 5 reports with same issue. We are making more investigation to get some clues and let you know asap. Thanks for your attention.

SergioFerrari
Active Contributor
0 Kudos

Just to be sure, may I suggest to use the simple tool I published some time ago in SDN to verify the speed of the two different server? [Simple ABAP Benchmark|http://www.sdn.sap.com/irj/scn/weblogs?blog=/pub/wlg/3024] [original link is broken] [original link is broken] [original link is broken];

In add, I had similar problem because of the table buffer: there was a large table defined as buffered 100% (complete) in the DDIC (SE11) and in one server the table buffer size was small. As a result there was a continuous table swap decreasing so much the performance.

It was solved defining the table as not bufferable.

Sergio

Former Member
0 Kudos

Hi Sergio, I've tested the two servers with your program and the result showed significant speed issue with dbci server. Please see the results as following:

Transaktion: SE38

Prozessor(s): 9000/800

Host: sscr3db

OS: HP-UX

DB: ORACLE

DB-Release: 10.2.0.4.0

SAP-Release: 620

Execution time per Unit: 30 seconds

-


DBCI DIALOG

CPU Power - Primes found: -


3.280 10225

File System - - Written Mbytes: -


100 2020

RDBMS update - Updated records: 6.700 12300

RAM - Table scans: -


4 75

I think the speed of RAM & File system is obviously abnormal. Maybe I should find some tool running in OS to test the system speed.

Edited by: guoqiang zhang on Jan 17, 2010 8:50 AM

Former Member
0 Kudos

It is caused by a unusual process in OS which occupied all CPUs. thanks for all your support.

SergioFerrari
Active Contributor
0 Kudos

Obviously the tool is very simple and measurement are very influenced by the current workload of the machine.

The CPU power should be measured when no other relevant process are running but normally, if you are not close to 100% utilization, it is quite constant.

File system is not so important, it is most related to interface programs reading/writing on sequential files.

RDBMS update is very important and normally it is important also to exclude network problems between DIALOG and DBCI. So your measurement is strange because normally the opposite occur. Probably the issue is well indicated by the last measurement.

RAM Table scan - I fill a large internal ABAP table into memory and then a scan it sequentially. It could be that DBCI RAM or some SAP buffer (ROLL Area,...) is too small and the scanning requires a lot of OS swap decreasing so much the speed.

I would verify OS paging, Oracle SGA size and allocation, ABAP ROLL area parameters and ABAP table buffer, ABAP SQL trace .

If other SAP instances are running in the same DBCI I would suggest to try a test shutting them down.

Sergio

Former Member
0 Kudos

Hello Sergio,

I'm trouble shooting an R/3 system , and encounting this post.

I would like to ask you what is the program you provide here for system scan?

Thank you very much.

BR,

Kate JI

Answers (2)

Answers (2)

markus_doehr2
Active Contributor
0 Kudos

Keep in mind that the database caches the results. So if you execute this the first time on the DBCI all data need to be read into the database cache, this takes those 9 minutes. If you then execute the same statement on the dialog instance, all necessary data is already in the cache so no more I/O necessary and hence it's much faster.

To really compare those two servers you'd need to

- shutdown the system including database

- startup

- execute on DBCI

- again shutdown everything including database

- start everything

- execute on dialog instance

Markus

Former Member
0 Kudos

Thanks Markus.

I believe this is not related with DB cache since it behaves constantly each time regardless where we run the report first.

Former Member
0 Kudos

Hi,

Check the network configuration between the dialog serverand the DBCI server.

Verify specially the speed and duplex setting of the ethernet interface.

If it is set to "auto" it can create the kind of problem you are experimenting.

Regards,

Olivier

Former Member
0 Kudos

Olivier, thanks for replying. I think the result would be the opposite if network configuration was the cause of the issue. Our situation is that dialog server is great but dbci is slow.