APO DP - use of parallelisation profiles

[Former Member]

I am using APO DP V5.

Has anyone any experience of using parallelisation profiles (/sapapo/sdp_par) to speed up job processing times?

What is the recommended number of parallel processes?

Thanks,

Bob Austin, Atos Origin

[Former Member · Accepted solution]

hi

Generally the max number is 3 with block size of each 1000.

you can select the parallal process profiles in the each respective job.

thx

venkat

[wilian_segatto]

Closing old thread.

[Former Member]

The settings also depend on the type of processing you are attempting to do, the following three SAP Notes give guidance and the "starting" settings you should try. It also explains the best way to tweak these settings:

a. SAP Note 961273 – Parallel Profile and SNP Heuristics

b. SAP Note 961483 – Parallel Profile and Deployment

c. SAP Note 853142 – Parallel Profile and DP Batch Runs

Regards

Ian

[Former Member]

Yes we have used it in our projects for improvement of processing time..

Depend on the data volume of each applications, we have maintaied the block size

Parallele processing is always useful for the huge data to be process in Background.

To improve the performance of such background jobs it is often advisable to run these jobs in several parallel processes

By using parallel processing you no longer have to split jobs into smaller units

Parallele Processing Profile Attributes:

If you see in the Process field max you can maintain 9999

Block Location Product :If you do not specify a block size, the program uses a block size of 20,000 as the default value.

e.g. of block size

You want to copy 13,000 location products and have chosen a block size of 5,000. The program starts three parallel work processes. 5,000 location products are copied in the first and second work processes, and 3,000 are copied in the third.

Block Size Peging area:

If you do not specify a block size, the program uses a block size of 10,000 pegging areas as the default value.

e.g.: You want to copy 13,000 pegging areas and have chosen a block size of 5,000. The program starts three parallel work processes. 5,000 pegging areas are copied in the first and second work processes, and 3,000 are copied in the third.

Block Size Orders:If you do not specify a block size, the system uses a block size of 500 orders as the default value.

Block Size Stock:If you do not specify a block size, the program uses a block size of 1000 stocks as the default value.

No Order Serialization:When orders are copied in parallel, serialization effects can occur in liveCache due to locks. This means that order posting can be executed one after another in liveCache due to locks, although the program has started the work processes in parallel.

If you set this indicator, serialization of orders does not occur, but the orders are actually processed in parallel. You can thereby improve performance when copying orders.

Block size : The number of elements processed in one process. The data objects that correspond to these elements differ from application to application

Maximum number of parallel processes :

Since one process is required for control purposes, the smallest entry in this field that results in performance improvements is 3.

You can assign a server group to the profile so that the system only uses particular servers parallel processing. Other servers can then be ‘reserved’ for users working online. You cannot create these server groups yourself; contact your system administration if you require new groups.

You can also define the block size. There are minimum and maximum value for each application in the system. If you make no entry, the system uses a default value that is defined for each application.

Note: It is advisable : Although it is technically possible, working with parallel processing with a dialog job(online execution) is not advisable. In this case the main process, which controls the parallel work processes, is a dialog process. For various technical reasons this can lead to a noticeable deterioration in performance.

[former_member407655]

Hi,

Revise OSS consulting note 961273. It has exactly what you need.

Make sure after activating parallel processing you monitor the load balance of the live-Cache.

[Former Member]

I think the answer depends on the number of work processes you have available on the application server. You should have your Basis team check this out. Once you know this, you can set this up as (number of work processes - 1), so that even if you are running a job with parallel processing, the system still has one work process open for other tasks.

We have a few PPPs that we use for different sized jobs

6 processes with block size 10,000

6 processes with block size 5,000

6 processes with block size 2,000

We apply them as appropriate to the size of the job.