Average workday throughput in January

In this subsection the workday (Mon - Fri) average values during January are displayed. The mean values are calculated at the periodical measurement times. More precisely, this means that a throughput measurement executed at hh:mm will be added to the mean value for hh:00. The mean values are averaged about twenty measurements. The following workday mean data are shown below.

In figure 7 the throughput workday mean values for the measurements between the ZAM host and the SARA and UU-36 hosts are presented. All hosts have an interface with a bandwidth of 100 Mbit/s. This plot is the averaged equivalent of figure 4, but only one day is displayed here.

  

Figure: This plot shows the throughput workday mean values between the ZAM host and the SARA and the UU-36 hosts. All these hosts have 100 Mbit/s interfaces. The data are obtained in January 2000. This plot is the averaged equivalent of figure 4. Show this plot in real size.

Figure 8 displays the throughput workday mean values between the IPP host and the FOM and UU-36 hosts. The IPP and FOM hosts have 10 Mbit/s, in contradiction to the 100 Mbit/s interface of UU-36. This plot is the averaged equivalent of figure 5.

  

Figure: This plot shows the throughput workday mean values between the IPP host and the FOM and UU-36 hosts. The first two hosts have 10 Mbit/s interfaces, the latter has a 100 Mbit/s interface. The data are obtained in January 2000. This plot is the averaged equivalent of figure 5. Show this plot in real size.

To make a comparison between the two DFN routes two Jülich and Stuttgart, figure 9 shows the throughput workday mean values between the FOM host and the IPP (Jülich) and RUS (Stuttgart) hosts. All hosts have 100 Mbit/s interfaces. This plot is the averaged equivalent of figure 6.

  

Figure: This plot displays the throughput workday mean values between the FOM host and the IPP and RUS hosts. All hosts have 10 Mbit/s interfaces. The data are obtained in January 2000. This plot is the averaged equivalent of figure 6. Show this plot in real size.

From these figures the following conclusions can be drawn:

1.

The performance decreases for the connection ZAM - UU-36 are in the plot with average values (figure 7) less deep than in the daily plot (figure 4). A part of the explanation is that in the beginning of January the performances during daytime were better. Another explanation is that at Mondays and Friday network performances seems to be better during daytime. The quarterly mean data for the days of the week will make this more clear. A third possible explanation is that the daily performance decreases are not always at the same time of day. This implies that the performance dip is broader, but not so deep. 

2.

The daily data plot of the connections IPP - FOM and IPP - UU-36 (figure 5) is in good agreement with the plot of the workday mean data (figure 8). However, in the latter plot the fluctuations are clearly averaged out. About the same explanations as in item 1 can be given here. 

3.

The same conclusions for the comparison between the daily data plot of the connections FOM - IPP and FOM - RUS (figure 6) and the workday plot (figure 9) as in item 2 can be drawn.

4.

The performance dip during daytime in the connection FOM >> RUS (figures 6 and 9) are also found in the other connections to the RUS host.