need help for deleting XI message performance header

[Former Member]

Hi,

In our production system, we are sheduling job SAP_XMB_PERF_REORG to removes all XI message performance headers (as per SAP note 820622). In this job we are using report SXMS_PF_REORG. But when are trying to execute this report, it asks for some input parameter.

Can anybody gives us detailed information about these parameters? Below are the 5 input parameters.Also please let us know whether all these parameters are mandatory or optional.

1. Days Until Raw Data Deletion

2. Days Until Delete Aggregation

3. Block Size

4. Max Data Record No. per Select

5. Parallelization Level

Thanks & Regards,

Sari

[former_member334189 · Accepted solution]

per 5: With this parameter you can activate the built-in parallelizer. This one is deactivated per default and it was implemented for dealing with giant backlogs. If, for example, the reorganization did not run for quite some time and there are millions of data records remaining in the database, the parallizer is designed to deal with it. However, activating the parallelizer will put a high load to the database and it will also consume a large percentage of CPU resources.

What happens in detail is this:when running this report as batch job having the parallelizer deactivated the job will simply use it's own background work process. When activating the parallelizer this background work process starts up several parallel tasks running in dialog work processes. This is pretty important as entire message processing is done in dialog work processes. Consequnently a large number of parallel reorg tasks will speedup the reorganization, but it will also impact the overall performance of message processing.

Enter -1 to make the system default 1 active (meaning: degree of paralleization = 1 = parallelizer off) or enter any value 2 .. 5 to activate the parallelizer and to run up to 5 reorg tasks in parallel.

As lined out above there is no need in the normal case for setting all these parameters explicitely. Simply run the report using the default -1.

In case you do want to use the parallelizer you should adjust the number of data records read per select (4.) to an accordingly. This number should at least be 10 times higher that the degree of parallelism (5.) multiplying by the block size (3.).

Example: you would like to run 4 reorg tasks in parallel and each task shall delete up to 6,000 records in one go.

Hence,

Max Data Record No. per Select >= 10 * 4 * 6,000 = 240,000

Here you might want to set Max Data Record No. per Select to 250,000.

Best regards,

Harald Keimer

XI Development Support

SAP AG, Walldorf

[Former Member]

Dear Harald,

Thanks a lot for your help. The way you have given explanation about each parameter is really appreciable.

Regards,

Sari

[former_member334189]

Dear Sari,

when you start the report the UI is coming up with default value -1 for all parameters. This value makes the subsequently invoked reports adhere to the system configuration. When scheduling the report to be executed as a batch job this default value is used automatically. Thus, scheduling the report using SM36 will do all the work in background. For common house-keeping there is no need to worry about the parameters.

So, what are these parameters good for?

Well, these parameters have been introduced to deal with extraordinary situations. Using these parameters one can override the system defaults. Thus, one should be rather careful when setting these parameters to values other than default -1.

Let's go into details:

per 1. & 2.: with these parameters the parameters PERF - DAYS_TO_KEEP_DATA - AGGREGATION_ITEMS and PERF - DAYS_TO_KEEP_DATA - MEASUREMENT_ITEMS are overridden.

per 4: this parameter controls the number of data records read from the leading table for performance data. Enter -1 to make system default 200,000 active or enter any value 1 .. 300,000.

per 3: Once a certain number of data records is read from leading table these records are deleted blockwise. Enter -1 to make system default 3,000 active or enter any value 1 .. 50,000. Keep in mind that deleting n leading records means that in addition m related records from another table are deleted also. The relation n:m varies from 1:0 to 1:30 (depending on the release, the active configuration and the scenario). Consequently deleting 50,000 leading records might trigger 1,5 mio related records to be deleted. This can lead to serious db related issue (for example, roll back segment overflow), thus, be careful!