Show TOC

Procedure documentationEvent Calculation: Configuration of Parallelization

 

The core of Technical Monitoring and Alerting is the Event Calculation Engine (ECE), which performs the most important MAI calculations. For general information about the ECE, see Optimizing the Performance of Event which describes how you can improve the performance of the ECE by optimizing your parallelization settings. There are three configuration parameters, which are explained below.

Parallelization can improve the performance, in particular of large system landscapes with a lot of managed objects, decisively, but the ECE occupies several of the dialog work processes of your system, at the same time.

Caution Caution

Caution: the parallelization settings affect the event calculation intimitely. The parameter values can cause significant performance and load differences,so you must observe the consequences of each change carefully. Search for current SAP notes about the “Event Calculation Engine” which may be relevant for the configuration settings.

End of the caution.

The parallelization of the ECE is controlled by the following parameters:

  • ECEDESTGRP

    This parameter specifies the name of an RFC server group. The ECE calculations are only performed on the instances of this group. You can, for example, reserve some instances for the ECE calculation. Create and edit these groups in the transaction RZ12 (CCMS: RFC Server Group Maintenance)

    The default value is the group SPACE, which contains all instances in the system, and always exists.

    Note Note

    Group names comprise only upper-case letters and numbers.

    End of the note.

  • MAXWPSPERC

    The number of work processes occupied by parallelization is limited by this parameter. The parameter specifies the maximum percentage of the dialog work processes which can be used by ECE. Only the dialog work processes of the instances of the specified RFC server group count (see parameter ECEDESTGRP).

    The default value is 20.

  • SUBGRPSIZE

    ECE first checks which calculations can run in parallel, i.e. which calculations do not depend on the results of other calculations. This is the case for different categories in different managed objects, which is referred to as an independent subnet. Complete parallelization (all independent subnets are calculated in parallel in their own work processes) is not desirable, because it would occupy too many work processes, and each calculation would require its own RFC. Subnets are grouped to reduce the number of RFCs required. This parameter specifies the target group size.

    The default value is 20.

Procedure

The default settings of the configuration parameters MAXWPSPERC and SUBGRPSIZE are probably not the optimal combination of values for you. The best combination of values depends on your circumstances, such as the number and type of managed objects, configured monitoring use cases, or the performance of your SAP Solution Manager systems, so you must determine this best combination yourself. Do this in three steps.

  1. Schedule variants of the ECE parameters, test various parameter combinations, and record their results

  2. Interpret the results

  3. Change the parameter values according to the results obtained

Note Note

This procedure will help you to determine the best settings of the parameters MAXWPSPERC and SUBGRPSIZE. Setz the RFC server group (parameter ECEDESTGRP) independently, to distribute the load in your system. .

End of the note.

Perform the following steps:

Scheduling an ECE parameter variant

To measure the performance of various parameter value combinations, proceed as follows:

  1. Start transaction SE38 and run the report ECE_CREATE_PERF_SCHEDULE.

  2. In the Change Subnet Groups group box, specify which area of parameter SUBGRPSIZE to check, and with which step size. Do not initially change the default settings (5-50 with a step length of 5).

  3. In the Change Processes Used group box, specify which area of parameter MAXWPSPERC to check, and with which step size. Do not initially change the default settings (10-50 with a step length of 10).

  4. In the Frequency of Change group box, specify for how long each combination is to be tested. The results log is retained for two weeks.

    Example Example

    If you use the default values for the subnet groups and percentage of the work processes, 10 x 5 = 50 variants must be checked. To test each variant for 6 hours would take 300 hours, or 12,5 days. At this time, just the results would be available.

    End of the example.

  5. In the At the End of Scheduling group box, you can specify whether to reset the parameter values to the default values (20/20) after the variant. This flag should be active.

Interpret the results

After testing with the report ECE_CREATE_PERF_SCHEDULE, interpret the results, i.e. determine the parameter value variant with the best ECE performance (this test takes somewhat more than two days with the default settings). Proceed as follows:

  1. Start transaction SE38 and run the report ECE_PERF_STAT.

  2. In the Select Function group box, choose Calculate Proposed Parameter, and set the Ignore Application Instance flag.

  3. The best combination of values of the parametersContaiGröß (SUBGRPSIZE) and %DialgoWP (MAXWPSPERC) are shown, for both the shortest run time and the highest throughput. A final recommendation is calculated from these values. Set the parameters to the recommended values.

If you are not sure thow good the recommendation is, you can schedule another ECE parameter variant, and concentrate on the recommended range with shorter steps. The results are retained for two weeks.

For a deeper interpretation of the results, choose Display Raw Statistics in the report. The following results are output each measured value combination of the two parameters:

Column Title

Description

Number

The number of measurements; the greater this number, the more reliable the results

Mean Run Time

Mean run time of ECE, in milliseconds

No. Subnets

Mean number of independent subnets

Resources

The mean number of resource bottlenecks; a bottleneck occurs when the ECE cannot use the number of work processes specified in the configuration, because there are too few work processes available. This value should be less than 0,5 on average.

COPS/Categ

Mean number of ECE calculation operations per second

Throughput

Mean throughput of ECE in metrics and rules, per second

Metric/Cat

Mean number of metrics calculated

Rules/Catg

Mean number of rules calculated

Changing parameter values

To set the parameters SUBGRPSIZE and MAXWPSPERC according to the recommendation, proceed as follows:

  1. Start transaction SE38 and run the report ECE_MAINTAIN_CONFPARAM.

  2. To change this parameter, enter the name in the Name field, and the recommended value in the Value field.

  3. Choose Execute, to change the value of the parameter.

Result

You have optimized the ECE parallelization settings.

The function and performance of ECE is also monitored and output in self-monitoring, in the technical component Event Calculation Engine (see Self-Monitoring of the Monitoring and Alerting Infrastructure)