The “Utilization real” report shows the CPU utilization monitored (also in VM2000 operation). The “Utilization normed” report should not be used, in particular not in VM2000 operation!
Assessment of the CPU workload varies according to the type of application and type of load. A workload of 100% is feasible and desirable for pure batch applications. Balanced utilization of the remaining resources is to be strived for.
In the case of mixed loading with TP, interactive and batch applications, the main application should comprise no more than 70% of the total CPU workload for the system (in the case of multiprocessor systems, depending on the number of CPUs, up to 90% is acceptable for the main application). An ideal background load triggers neither I/O operations nor paging; being extremely compute-intensive it is well-suited to making use of surplus capacity. This combination of features makes it possible to take full advantage of CPU capacity; in practice, however, it seldom occurs. Any one I/O operation or paging transfer can interfere with the TP and interactive applications. I/O-intensive loads and programs with extensive memory requirements should not be loaded in the background. The background load must be made clearly subordinate to the main application by assigning appropriate priorities or PCS settings. Ideally, the files of the main application should reside on other disk drives (and at least on other volumes) and be accessible via other paths.
The SIH share depends, among other things, on the number of I/O operations of all kinds and the intensity of task communication. The SIH portion should not exceed 20% (/390 servers) or 10% (x86 servers). A high percentage of TPR operations is caused by SVC calls, which perform resource-intensive system services. The workload can be regarded as normal if the following formula applies:
TU + TPR > 3 * SIH (/390 server) or TU + TPR > 7 * SIH (x86 server)
Now and again there are instances where the load violates the above rules without actually impairing performance behavior. As long as the response times are good there is no cause for intervention in these cases.
Generally the applications run on an x86 server in /390 mode. As programs which run in this mode have a higher CPU requirement than on /390 servers, a high share in the processor state TU is normal (some 60%, standardized as 100% utilization). If a higher standardized TU share is shown, this means the performance of the server with x86 architecture is below the nominal value (the official RPF value). However, of the (standardized) TU share is lower, this means the performance is above the nominal value.