Performance after Unicode Conversion

[Former Member]

Hi

I am in the middle of testing a migration from non-unicode to unicode. The migration seems to go fine and so far we have not found any problems with the functionality on the Test Unicode converted system. Our Response times in ST02, ST03 and DB02 are all very good. Although, when certain custom transactions and reports are run they are taking 10x's longer or more and in many cases timing out. There are many custom (Z*) reports and we don't think we should have to make code changes to all of these to make everything run as it did on the non-unicode system.

How can I debug this problem further to figure out where the issue lies?

I have traced one of the reports and found that a select statement is taking for every in the new unicode converted system but running very fast in the non-unicode system. 5.5 seconds in the non-unicode system vs. timing out in the unicode system

Here is the select statement that I seem to be getting stuck on.

SELECT vgbel vgpos vbeln posnr matnr lfimg FROM lips INTO TABLE it_lips

FOR ALL ENTRIES IN it_vbak

WHERE vgbel = it_vbak-vbeln.

Any ideas?

Lee

[nils_buerckel]

Hi Lee,

did you adapt profile parameters (for memory) as listed in SAP note 790099 after the conversion ?

Best regards,

Nils Buerckel

[Former Member]

Yes. Profile parameters were all included. Additional information found today is that many of the indexes were not loaded after the conversion. From talking to some of my SAP DBA counterparts, it turns out that if you make your indexes at the DB level instead of in SAP, you will loose them when you perform the migration. Indexes need to be stored in the data dictionary and that only occurs when you create them within SAP.

So, now I am trying to figure out how to get a list of all the indexes that were created at the DB level, so we can recreate them within SAP.

Anyone know the best way to do that?

Lee

[nils_buerckel]

Hi Lee,

in fact this is true: the system copy procedure used for the Unicode conversion takes care of all tables included in the ABAP dictionary. Any tables created on DB level only by customers or partner solutions will not be created automatically on the Unicode target system.

In your case maybe a simple compare between Non-Unicode (e.g. PRD system) and Unicode sandbox system on DB level should result in a delta list of missing tables / indixes ...

Best regards,

Nils Buerckel

[Former Member]

I did a comparison of the non-unicode system and found about 20 indexes that did not get copied with system copy. I've created most of them in SAP, there are six that I had to create at the DB level and we are investigating the right way to make those in SAP as well so this doesn't happen again.

Thanks for your responses.

Lee