Reduce the runtime of CDCLS table

[Former Member]

Hi,

We are in process of doing updrade with unicode conversion and migration.

We need to improve the performance for the large tables. The approach which we have used is given below:

- CDCLS is 44 GB ,

- Split into 32 packages during export

- Used latest R3load

- Used parameter settings for oracle

- Parallel processes run were 32

- 16 CPU , 16 GB RAM .

- Used Note 1043380 - Efficient Table Splitting for Oracle Databases for table splitting

- Export of CDCLS is running more than 72 hours

Please suggest how can we reduce the runtime of this table.

Kind Regards,

Srinivas Reddy

[Former Member]

Hey Srinivas,

1) Only splitting is not enough for accelarating the table unload. You will also have to arrange the packages in orderby.txt in such a way that you get the maximum thruput.

2) if you have CDCLS as 44 G, whats the total DB size and top tables in DB - this also matters how you split other tables and arrange in orderby

3) 16 CPU - is the CPU duel core? if yes you can have more than 32 R3load process

4) 32 splitts for 44 GB - as per my experiance this is too much. You definatly have problems with the thruput

5) I support Haralad for data archiving - this will give yougreat value add.

Thanks,

Amit Thombare

[Former Member]

Hi Srinivas,

I don't think you have more options that table splitting... You might can tweak Oracle settings to increase the performance, but it will take a long time for your export anyway..

What you could do (but it is not a cheap option) is performing a CBU (Customer Based Upgrade), where SAP is taking care of your process. This way a lot of things can be done in advance and in parallel to your normal operation..

Kind regards,

Mark

[Former Member]

I suspect the usual comment of avoiding data is out of question, but just in case you haven't thought about archiving, you might want to consider it (I think it's archiving object CHANGEDOCU). This might also help in your future system to improve performance...