z/OS MVS Programming: Workload Management Services
Previous topic | Next topic | Contents | Index | Contact z/OS | Library | PDF


Example

z/OS MVS Programming: Workload Management Services
SC34-2663-00

Example

In this example, there are three goal-mode systems in the sysplex, all running at 100% CPU utilization. Each system has two DNS servers running for a particular application. To assign weights to the three systems, WLM scans the CPU consumption table from the bottom up, looking for a level where at least one system has 5% or greater cumulative CPU consumption. In this case, Level 5 row is used because System A exceeds the 5% minimum (and so does System B). Level 5 is then used to calculate the weight for all systems including System C. So in this case the system weights are determined by the amount of importance 5 and discretionary work each system is running. Because there are two servers on each system, the server weight is the system weight divided by 2.

System weightServer weight
System A weight = 200 *64 / 640 = 20Server weight = 20 / 2 = 10
System B weight = 400 *64 / 640 = 40Server weight = 40 / 2 = 20
System C weight = 40 *64 / 640 = 4Server weight = 4 / 2 = 2
Table 17. CPU consumption table
System ASystem BSystem C
LevelSUs%SUs%SUs%
0200010020001002000100
11800 901900 951840 92
21600 801500 751600 80
31100 551500 75 800 40
4 400 201200 60 800 40
5 200 10 400 20 40 2
6 80 4 20 1 0 0
7 0 0 0 0 0 0

Starting with z/OS® V1R7 the IWMSRSRS service supports the function code SPECIFIC, that allows for calculated weights to also take the server-specific information into account. If the IWMSRSRS service is called with FUNCTION=SPECIFIC, then the weight calculated as described above, is additionally multiplied by a Performance Index (PI) factor and a queue time factor. The value of these two factors is calculated as follows:

The PI factor
is 1 if the server has a PI <= 1 which means that it achieves its goals. The PI factor will become < 1, if the PI indicates, that the server does not achieve its goal.
The queue time factor
becomes < 1 if the server is the owner of enclaves and those enclaves have reported queue times relative to their total elapsed times. This means that their arrival time differs from the enclaves create time, when the enclave was created.

Since z/OS V1R8 a third factor is multiplied into the weight calculation, the health factor. The health factor is set for the server address space either by the registration service IWMSRSRG or by the health service IWM4HLTH. For further information about these services, refer to IWMSRSRG – Register a Server for Sysplex Routing and to IWM4HLTH – Setting Server Health Indicator.

Go to the previous page Go to the next page




Copyright IBM Corporation 1990, 2014