Conclusions

The examples discussed in the subsections 3.2 and 3.3 lead to the following general conclusions:

• The RTPL package gives a reasonable insight into the performance of a network from a user point of view without heavily processor and network loads. Long term statistics may help to average out fluctuations, at least when the network is stable.
• The Netperf package gives a reasonable direct insight into the available throughput of a network for the end user. Average roundtrip and roundtrip package lost information also give some insight into it, however, to make it valuable a more advanced evaluation of the results as a.o used in [1] is required. However, we have the impression that these methods are always less direct than netperf measurements. Therefore, at least for short term measurements, we prefer the Netperf package.
• At the IPP - FOM setup (subsection 3.3) a mixture is used of sites with 100 Mbit/s and 10 Mbit/s interfaces. In the sequel we prefer tests with only 100 Mbit/s interfaces, because they give more clear information of performance dips by the results of their throughput measurements.
• In the IPP - FOM setup (subsection 3.3) we have the impression that more direct information at the TEN-155 backbone would be valuable to make the interpretation of the results still more clear. We understand very well that workstations at the backbone have a limited capacity for extra load. Therefore, in subsection 4.2 we propose a setup which is intended to limit the load as much as possible.