Performance Management for System i

Executive Summary, Report 150
The PM eServer iSeries Executive Summary Report shows an overview of the current performance situation and gives a forecast for the next five months. The colors that appear in the traffic lights depend on the capacity utilization of the iSeries server when compared to the guidelines that have been set for the system. A green light indicates that capacity utilization is, or is predicted to be, well below the recommended guideline; in other words, that the status is (Acceptable). You see a yellow light if the utilization is, or is predicted to be approaching the guideline, (Marginal, intermittent problems and you should plan action). A red light indicates that the guideline was reached or surpassed, or is predicted to be reached or surpassed (Critical, problems are almost certainly occurring and you will need to take action urgently).

All executive summary graphs are based on the average utilization during the two heaviest hours (peak average) for each business day during the measured period. Because the calculations are based on averages, the utilization during shorter periods may be higher or lower than the colors indicated.

The traffic lights represent:

• Processor - Interactive capacity
• 
  The interactive CPU guideline for each CPU model varies depending on the number of processors that are on the system. The recommended guideline for your AS/400 system is indicated on Report 650. A green light indicates that the utilization is below the guideline. A yellow light appears when the utilization is within 10% to 15% or less from the guideline based on the number of processors on the system or included in the logical partition. A red light appears when the guideline is exceeded. A red light means that the resource is constrained.
  
• Processor - System + Interactive
• 
  While the interactive capacity of a system or partition governs the amount of interactive work which can be done by the system another key factor which affects response time is the amount of  resource used by the system in addition to the interactive jobs. Typically when this value approaches the guidelines intermittent performance problems will be experienced.
  
• Processor - Total
• 
  Although it is possible to use 100% of the CPU, the threshold value for total CPU utilization is between 94% and 99% depending on the number of processors. This has been done to give you an indication of when batch job problems may occur. When the CPU utilization is close to 100%, the throughput for batch jobs may be very slow and the number of jobs on the batch job queues may increase. A red light appears if the total processor utilization is at or above threshold. A yellow light indicates that the processor utilization is within 5% to 10% of the threshold, based on the number of processors on the system or included in the logical partition. Anything below that is shown by a green light.
• Disk space
• 
  The default guideline for disk space utilization is 80%. Your current guideline is indicated on Report 850. A red light indicates that the guideline was reached or exceeded. A yellow light indicates that 1% to 10% is available before the guideline is reached. A green light indicates that more than 10% is available before the guideline is reached.

Sample report:

• When a resource is predicted to become constrained due to peak workload and/or due to average workload conditions.
• How long it will take to move from acceptable status to critical status. The size of the marginal (yellow) bar is dependent on your growth rate. The marginal bar is designed to allow for three to six months of growth, depending on the growth rate, before the particular resource becomes constrained or in critical status. During this time you should be planning for the resolution of the problem.

Sample report:

There are four components to the Management Summary Graph. They each address a component of the system which can impact performance and which will require action to correct. There are several alternatives:

• You can offload or reschedule work to reduce the load on the marginal or critical component
• You can tune the system and/or applications to reduce the demand for that component
• You can upgrade the constrained component for example, if disk capacity is inadequate and impacting performance then purchasing additional disk will relieve the constraint.

It is important to understand what is being represented on each graph.

Disk Space:

Color is used to highlight the status of disk usage over the period of the graph as follows:

These values are used because experience has shown they are the points at which typical installations start to experience performance problems because of disk constraint. They allow a three month lead time in which to correct the situation.

The figures to the left of the bars show the averages for the period for which the most recent data is available which is the first position of the bar. The rest of the bar shows the predictions for the following 12 month period.

In the case of a system implementing LPAR each partition collects DISK statistics only for that partition and the reports reflect the DISK characteristics for that partition and not the total system.

Processor - Total System Utilization:

In the current implementation of AS/400 CPU cycles are classified into one of three categories:

I = Interactive (Jobtype = I)
B = Batch (Jobtype = B)
S = System

The total amount of CPU cycles used by the system is the total of these three categories. Again, the figures to the left indicate the averages for the most recent period and the color of the bars represent the status of utilization based on "acceptable, marginal and critical" criteria. However, the values at which the colors change  vary depending on the number of processors on the system and are based on the guidelines for good performance and standard queuing models for n-way processors.

In the case of a system implementing LPAR please refer to the discussion on LPAR reporting as the reports show different and in some case misleading results depending on which release of OS/400 is being used. This caveat applies to all Processor Utilization reports.

The following tables depict the guidelines that are automatically considered when graphing your PM eServer iSeries data:

Processor - System + Interactive:
If the total of the System + Interactive cycles exceed the values in the following table interactive response time will usually deteriorate. Again this depends on the number of processors on the system.

  However, some systems such as the 7XX and 8XX and the server models have other constraints on the interactive performance and these bars can how an acceptable utilization yet interactive performance can still prove unsatisfactory. To understand how these systems are behaving it is necessary to consider the utilization of the Interactive Capacity.

Processor - Interactive Capacity:
The Server models of the AS/400 (models 50S, 53S for example) have much lower thresholds for Interactive work than the above chart would indicate. Likewise, the 7XX and 8XX models have Interactive Feature cards which provide interactive processing capability. The top chart on Report 175 shows the percentage to which the Interactive Capacity of these systems is utilized and the colors used to represent the status is shown in the following table.

   It is important to understand that the value on the left of the bar is expressed as a percentage on the Interactive Capacity of the system and not the total processing capacity of the system. The reason this is done is to show what percentage of the component is being used as it the constraint of this and not the total system which indicates if action is required.

   For AS/400 non-server models and those which do not implement the Interactive Feature card(s) the values shown on this chart are those used for Interactive work (Jobtype = I) and these are expressed as a percentage of the total processing capability of the system.

   Processor - Summary:
In order to understand the full performance picture and the actions you need to take  it is necessary to look at the three sets of bars together. The following is a generalization and you should remember you can always relieve a performance bottleneck by offloading work or improving the tuning of the system or application.

1.   If the Processor - Total is a problem you need more basic processor power and a processor upgrade in indicated. This may mean a complete system model upgrade or simply a feature upgrade within a model. It could mean a higher powered processor which will  reduce the utilization percentage but it could also mean more processors which will in turn change the percentages used for the "Acceptable, Marginal, Critical" criteria. It is important to understand the upgrade options and the impact they will have on the "# of Processors" charts above.
2.   If the Processor - System + Interactive is highlighted then again, you need more processing power and an upgrade is indicated.
3.   If the Processor - Interactive Capacity is indicated as a problem then you clearly need to increase the Interactive Capacity of the system. In the case of a Server model this means a model upgrade or a system upgrade. For a 7XX or 8XX it may be possible to simply install an additional Interactive Feature Card (or cards) but if you are already at the maximum for that model a system model upgrade is indicated.