In order to ensure comparability of the results, there were
also measurements for the actual load during the execution.
These measurements indicated that the actual load is very
close to what was actually defined. And the median numbers
are equal to what was defined. Similar kinds of measure-
ments were done with the relations between different kind
of scenarios. In that case too, the actual relation of scenarios
is very close to what was defined. What this all means is
that the differences in workloads are not significant in these
performance measurements.
The CPU load and memory consumption were measured
during the test set execution. In both systems in IMS and
SIP the CPU load is very similar, and even the maximum
and minimum values are almost equal. In both systems, the
CPU load starts around 18% in the first step (200 scenar-
ios/second) ending up close to 50% with the full load of 600
scenarios/second. With the memory consumption, the situation
is different. IMS systems consume significantly more memory
than SIP systems. In the fist load step, the IMS system had
4954 MB of memory free and in the final load step of load
4206 MB. That’s roughly 750 of MB memory usage. In the
SIP system the overall memory consumption is around 100 MB so it is significantly less. There is also a difference in
the way how memory is used. In the SIP system the memory
is, to some extent used already during the first step and there
is only a slight increase in memory usage during any further
test execution. In the IMS system, the memory consumption
follows closely the load of the system.