Mass Assignment of Organization Sets with Manual Price Plans
System customizing/settings
The new functions are based on the principle of processing parallelization. That means that the response time is heavily influenced by the number of system resources that are available, especially the number of CPUs and the number of available dialog work processes. This assumes that many free resources are normally available for use on the system.
A server group, which dynamically distributes the workload among all allocated active instances, is used for parallel processing. The server group is defined by the system administrator in transaction RZ12.
To avoid overloading an instance, corresponding instance parameters must be set correctly. These settings must be made in the system profile and essentially relate to the number of dialog work processes available and the number of CPUs.
The current settings can be displayed using transaction RZ12 in the detailed view of the server group, and, provided that the user has the required authorization, they can be changed temporarily. However, this change is only valid until the next restart of the instance. Final changes must be made in the instance profile.
Note
Keep in mind that this setting always applies for each instance, independent of the server group that is used to address it.
You cannot specify a generally valid optimal value for the number of tasks running in parallel. As a rule of thumb, when you are using one central system (that is, when the database and exactly one instance being used are on the same system), you can assume that one task per CPU is an appropriate value. When using a distributed landscape (that is, when the instances being used are running on other systems to the database), you can schedule approximately 1.5 tasks per CPU of the instance.
You usually control the number of parallel tasks that are running at the same time on one instance by setting the minimum number of free work processes (parameter:
Minimum Number of Free WPs
). However, you must also consider the limits for the other parameters in order to calculate the actual limit for the number of parallel tasks.
Note
Note that these settings only represent upper limits. If an instance is already temporarily busy with other active processes, it may be possible that the system cannot make the planned number of parallel tasks available.
The new function of the Price Planning Workbench (PPW) not only processes parallel tasks; it also runs processing in the calling task at the same time. If possible, you should consider this when setting the server group parameter.
When using a central system, you determine the correct value of the instance parameter by using the following formula:
Minimum number of free WPs = Number of available dialog processes – Number of CPUs + 1
Note
Note that this reduction of the tasks running in parallel by 1, due to the active root task, also applies when another application, in which the root task does not cause a workload, uses the instance for parallel processing.
In a distributed landscape, the instance in which the user is logged on, meaning, the instance on which the root task runs, is not normally specified. In this case, it makes less sense to consider the root task when determining the upper limit.
You can determine the number of CPUs of an application server by starting transaction ST06.You can see the number of CPUs in the upper right area of the list created by the transaction.
Significantly more than a total of 10 CPUs are normally used in a retail production system.
You can check the number of dialog processes on an instance with transaction SM50 (Process Overview). The transaction creates a list of all available system processes. You can determine the number of available dialog processes from the number of all rows in this list that contain the string
DIA
in the
Type
column.
Note
If too many parallel processes are started on one instance, they may all run more slowly, due to low availability of the CPU, because they have to wait for each other. In the worst case, this may even have a negative impact on performance.
Customizing setting PPW
The main customizing (
General parameters)
of the PPW has been supplemented by 2 additional parameters:
Server group
and
Limit for parallel processing.
Limit for parallel processing
: You set the utilization of parallel processes using this parameter. The number that you enter is the minimum number of price activation levels that are processed together in one parallel process. You can optimize performance by changing this parameter.
You change these settings in the Implementation Guide (IMG) in
Logistics – General -> Price Planning -> Configure General Parameters.
Setting the limit value for parallel processing
Because the new functions were chiefly developed for large price plans, you should use a large price plan to set the limit value for parallel processing.
In order to determine the number of price activation (PA) levels that need to be calculated, multiply the number of articles by the number of PA levels of the organization set for assignment and divide this value by the number of CPUs. Use the whole number portion of the result for the limit value for parallel processing.
Now determine the system response time by using the new mass assignment function.
Next, undo the assignment by deleting all articles of the price plan and then copying them (without the default organization set); save the price plan once more.
To fine-tune the settings, you can adjust the limit value you have set up or down. You should use the setting that leads to the shortest response time.
You should make this one-time setting at a time when the system is not being used by other users; this is necessary because you would not be able to compare the measured values taken at times that the system is being used by others, since at such times the system changes the number of parallel processes available, depending on its current load.
Refer to Mass Assignment of Organization Sets for information about performing assignment and to Performing Mass Price Changes for information regarding mass price changes.