System Copy - Split STR Files

[Former Member]

Dear friends,

We are planning to do system copy for SAP system based on SAP NW7.01 with oracle with database size of around 1.4TB.

and I want to use Table Splitting to optimize downtime..

I have selected 6-10 largest table of size more than 20GB and want to start with it.

For step "Preparation for Table Splitting" i have created a file "tabsplt.txt" with table names & no of splits as mentioned in system copy guide"System Copy for SAP Systems Based on SAP NetWeaver 7.0 Including Enhancement Package 1 ABAP". (by running this step, i can see it has created *.WHR files for mentioned tables).

For step "Database Instance Export" I have a screen "SAP System > Split STR Files".. i need your help.

1.   Largest tables in seperate package          >>What this means ?

     Number of table to be extracted               >> What is recommended ??

2.   Extract table with size more than limit    >>What this means ?

     Table size limit[mb]                                 >> What is recommended Table size limit??

3.  Split package with size more than limit    >>What this means ?

     Package size limit[mb]                             >> What is recommended Package size limit??

4.   Split predefined tables                            >> What this means ?

     Table input file*                                       >> Should this be same file what i have used for table splitting or something else ??

I am not finding much details in System copy guide for this.. if you know any document explaining same, it will be great.

Please let me know your recommendations & help.

- Best Regards

[paul_power · Accepted solution]

Hi,

I try to answer your queries: This is not a particularly vital step, the values are more based on your preference and can play around with them to test which gives you the system copy perfromance you want on a ttest system copy before doing ont he actual one:

For step "Database Instance Export" I have a screen "SAP System > Split STR Files".. i need your help.

1.   Largest tables in seperate package          >>What this means ?

Do you want the large tables (you mention 6-10) to have their owen package or be included with other tables in a packager. Its more so you can see that the package for <large table1> exactly corresponds to <large table1> and can see exactly how long thast table/package took to import. Not having to factor in that other smaller tables were nicluded.

     Number of table to be extracted               >> What is recommended ??

The number of tables

2.   Extract table with size more than limit    >>What this means ?

A table of size larger than <limit you enter> is split

     Table size limit[mb]                                 >> What is recommended Table size limit??

This is the limit where it will split the table if it is beyond the size you set. There is no recommended size, it depends on what you would consider a large table. You mention yourself that you have select 6-10 tables of size greater than 20GB

3.  Split package with size more than limit    >>What this means ?

If its extracting a taqble larger than 20Gb then this could make a very large single split package. It is usually preferable to split this packag into smaller "chunks". One reason is for performance, say if you are importing a large table and have configured the import to use 3 parrallell imports. You would likely prefer to import this table as soon as possible so its best to split it into small chunks so intead of a single import process working on a very large package, the number of parralllell import jobs you have (eg 3 -5) can each be working on a a split of this large package so the overal package gets import quicker.

Its also usefull in case a fault occurs, at least you will see how many smaller splits of a package have sucessfully imported and can pick up with the next small split of a package should an error occur during the import.

     Package size limit[mb]                             >> What is recommended Package size limit??

Again, there is no recommended limit. I would say as a very general rule, try to split the large package so that you end up with about 2 or 3 times the number of parrallell import processes you are going to use. This way each package should not be very large.

4.   Split predefined tables                            >> What this means ?

In case you dont want to consider the table size, but you know the names of certain tables you wish to have split, regardless of their size. You can use this.

     Table input file*                                       >> Should this be same file what i have used for table splitting or something else ??

This is different, its where you would create a text file which lists the tables by name which you desire to have split

Hope this helps.

Regards,

Paul

[Former Member · Accepted solution]

Thanks Paul.

But based on your answers, I have more questions on them..

1.   Largest tables in seperate package          >>What this means ?

Do you want the large tables (you mention 6-10) to have their owen package or be included with other tables in a packager. Its more so you can see that the package for <large table1> exactly corresponds to <large table1> and can see exactly how long thast table/package took to import. Not having to factor in that other smaller tables were nicluded.

     Number of table to be extracted               >> What is recommended ??

The number of tables

• >>> So this is only for keeping large tables in seperate package.. so how it will keep additional package ? for example, if i give Number of table to be extracted = 25. so, in this case, it will create 25 additional package for LARGEST tables ??

• >>> And, i understand, its not going to improve performance much ??

2.   Extract table with size more than limit    >>What this means ?

A table of size larger than <limit you enter> is split

     Table size limit[mb]                                 >> What is recommended Table size limit??

This is the limit where it will split the table if it is beyond the size you set. There is no recommended size, it depends on what you would consider a large table. You mention yourself that you have select 6-10 tables of size greater than 20GB

• >>> Is this a simple table splitting based on size ? If yes, then how its diefferent from step "Preparation for Table Splitting" in which I have created a file "tabsplt.txt" with table names & no of splits ?? and why we need step "Preparation for Table Splitting" ??
• >>> And how many splits it will create ? (bcos in file "tabsplt.txt", I also have to mention no of splits)
  

4.   Split predefined tables                            >> What this means ?

In case you dont want to consider the table size, but you know the names of certain tables you wish to have split, regardless of their size. You can use this.

     Table input file*                                       >> Should this be same file what i have used for table splitting or something else ??

This is different, its where you would create a text file which lists the tables by name which you desire to have split

• >>> Is this just a alternative of point 2 where i can provide a name of tables(for example may be 25 largest table observed in DB02 transaction)
• >>> And how many splits it will create ?

If you know any document explaining same, it will be great.

Thank you.

- Best Regards

[paul_power · Accepted solution]

Hi,

This thread gives a good detailed explanation:

http://scn.sap.com/thread/2088749

in addition to this,

package split is in addition to the table extaraction. It means that for example if table MARA was huge, it would create a large package. You might want to split this package into say 20 parts so that parallel import jobs could each take a part of it and perform the import in parrallel so it is done faster.

Split pre-defined tables is an alternative, it is examining table names not the size, the table size method focuses on the size of the tables, doesnt matter what tablename it is.

Let me know if further questions persist.

Regards,

Paul

[Former Member · Accepted solution]

Hi,

2.   Extract table with size more than limit    >>What this means ?

     Table size limit[mb]                                 >> What is recommended Table size limit??

You can't  put table size limit beyond 1.78 GB...Read below example:-

A customer database of an ABAP SAP System has 10 very large tables that are between 2and 20 GB in size and some other large tables ranging from 500 – 2.000 MB. After the JAVA- or Perl-based Package Splitter was executed with option “-top 10” (move the 10 largest tables to separate *.STR files) 10 additional *.STR files exit, but contain other tables than expected.What can be the reason of this behavior? What happens to large tables?

R3SZCHK limits the computed table sizes to a maximum of 1.78 GB. Because of this, the package splitter catches the first 10 largest tables found in the *.EXT files. A 20 GB table will have the same *.EXT entry as a 2000 MB table

Regards

S. Basis