Siemens SM2 monitoring system 用户手册
U3585-J-Z125-8-76
345
Notes on SM2 operation
Accuracy of the SM2 data
6.2.2
Notes on particular variables
CPU utilization
The term “percentage deviation” used in the following always denotes the absolute
deviation, in other words:
X and Y have a deviation of P if abs (X-Y)
deviation, in other words:
X and Y have a deviation of P if abs (X-Y)
≤
P.
TU/TPR/SIH time, IDLE time, or /390 time with SR2000.
The operating system uses timers to determine three time components: the total for TU and
TPR times, the SIH time and the IDLE time. In the case of SR2000, the /390 time is also
determined as a fourth time component. Consequently, falsifications due to the non-inter-
ruptible SIH code cannot occur. Investigations have shown that this timer is sufficiently
precise for realistic monitoring cycles. As a rule, comparative measurements made with
hardware monitors under typical load conditions have shown deviations of less than 1%.
TPR times, the SIH time and the IDLE time. In the case of SR2000, the /390 time is also
determined as a fourth time component. Consequently, falsifications due to the non-inter-
ruptible SIH code cannot occur. Investigations have shown that this timer is sufficiently
precise for realistic monitoring cycles. As a rule, comparative measurements made with
hardware monitors under typical load conditions have shown deviations of less than 1%.
SM2 divides the total for TU and TPR times into separate values for the TU time and the
TPR time on the basis of samples taken during the sampling period. The accuracy of this
procedure depends on the number of samples, which means that large deviations are
possible if few samples are taken. The /390 time is also affected by this inaccuracy, because
it is output as a relative proportion of TU.
TPR time on the basis of samples taken during the sampling period. The accuracy of this
procedure depends on the number of samples, which means that large deviations are
possible if few samples are taken. The /390 time is also affected by this inaccuracy, because
it is output as a relative proportion of TU.
Device utilization
As regards the hardware, the device is busy from the time an I/O request is received until
data transfer terminates. As regards the software, the device is busy from the time an I/O
request arrives in the I/O system module until channel termination.
data transfer terminates. As regards the software, the device is busy from the time an I/O
request arrives in the I/O system module until channel termination.
The values determined by the software monitor should therefore always be higher than the
actual device utilization. However, this systematic monitoring error is virtually balanced out
by the sampling delays caused primarily by the non-interruptible system states.
Comparative measurements with hardware monitors for typical load conditions and about
6000 samples have shown that the maximum deviation is
actual device utilization. However, this systematic monitoring error is virtually balanced out
by the sampling delays caused primarily by the non-interruptible system states.
Comparative measurements with hardware monitors for typical load conditions and about
6000 samples have shown that the maximum deviation is
±
2%.
Channel utilization
SM2 uses a special machine instruction to determine whether a channel is active or not.
Thus there are no deviations between hardware and software monitors in the interpretation
of this condition (see also restrictions on the ability of monitoring the channel load,
description of the CHANNEL report). The sampling delays caused primarily by non-inter-
ruptible system states result in the values determined by SM2 always being smaller than
the actual values. Comparative measurements with a hardware monitor for typical load
conditions and about 6000 samples have shown that the deviation is less than 2%.
Thus there are no deviations between hardware and software monitors in the interpretation
of this condition (see also restrictions on the ability of monitoring the channel load,
description of the CHANNEL report). The sampling delays caused primarily by non-inter-
ruptible system states result in the values determined by SM2 always being smaller than
the actual values. Comparative measurements with a hardware monitor for typical load
conditions and about 6000 samples have shown that the deviation is less than 2%.