Siemens Network Hardware SM2 monitoring system User Manual

U3585-J-Z125-8-76

349

Notes on SM2 operation

Even if system utilization is 100% (Hypervisor-Idle=0%), it makes sense for SM2 to provide
an IDLE value for the following reasons:

Full utilization of system load can be the result of a low-priority guest system which has
highly intense CPU time (in the extreme case: CPU loop), the system always obtaining
control as soon as the other guest systems voluntarily revert to IDLE mode.

The SM2 IDLE value permits the operator of a high-priority guest system to estimate the
extent to which additional load can be utilized. In this case the lower-priority system is idle
to the appropriate degree specified.

The general rule is: dilation resulting from other guest systems is always included in all SM2
monitoring values formed by the difference between two time values. This includes the
following monitoring information:

–

all I/O times for the monitoring programs DISK, TASK, SYSSTAT and FILE

–

response times for the monitoring program RESPONSETIME

–

catalog access times for the monitoring program CMS

–

QUEUE values in the task statistics

–

QUEUE values for queues Q1 to Q13 for the QUEUE-TRANSITION operand

Q0 (CPU-QUEUE) is without dilation, as the real CPU requirements from the product TU +
TPR are output!

The monitoring values supplied by the SERVICETIME monitoring program for systems
which use DCS, namely the values

–

DEVICE DISCONNECT TIME

–

DEVICE CONNECT TIME

–

FUNCTION PENDING TIME

represent the real hardware times (determined by DMS) without dilation by other guest
systems. Dilation of I/O times is reflected in the (REMAINING) SERVICETIME value.

All SM2 monitored data with “per second” (1/s) units refers to the elapsed time, and not to
the time assigned by the hypervisor for a guest system. As a result, the values determined
are less than those which actually occur during the time the guest system was active (for
example I/Os/sec. per disk, paging rate).

Disk utilization is determined by the sampling method while the corresponding guest
system is still active, in other words the values are not corrupted by other guest systems (in
contrast, the number of I/O operations does not refer to the active time per second, rather
to the seconds elapsed, and therefore the value appears to be lower).

In the case of the channel utilization values, the overall utilization per channel is output. The
percentage load caused by any one guest system cannot be determined here; as a conse-
quence the NON OVERLAP value (CPU and CHANNEL simultaneously active) has no
significance when an individual guest system is involved.