|Novel Approaches to the Monitoring of Computer Networks|
|Prev||Chapter 5. Tracking Network Growth||Next|
During the course of the experiment, requests were made that the data gathered was made available to people. To facilitate this, a web-based interface to the round robin database was developed. This front-end allowed the user to generate a number of graphs to compare the number of hosts on any given subnet over a particular time interval.
Like the graphs of Section 3.3, the graphs in the application were generated using Tobias Oetiker's rrdgraph(1) graphing engine. The graphs produced by this application, however, are significantly more complex to generate than those of Chapter 3 since they allow the user to choose the time interval over which they are generated, and to overlay information about various subnets onto the same graph. The advantage of this system is that it easily allows the user to compare different subnets.
Figure 5-1 shows how the application can be used to provide a comparison between three different subnets — in this instance, the 220.127.116.11/21, 18.104.22.168/21 and 22.214.171.124/21 subnets. It also illustrates the method used to determine the colour  of each line on the graph.
Using this web-based application, a number of interesting patterns in the usage of computers at Rhodes was discovered. Figure 5-1 shows one such pattern. Unlike the other two subnets shown, usage of 126.96.36.199/21 subnet clearly shows a tendency by some people to only have computers switched on during working hours. The computers on this subnet that are shown to be always switched on are probably those computers in the public computer laboratories, which are available to students twenty-four hours a day. The low peak in the graph on Tuesday 24 September can be simply explained and is interesting in itself; it was Heritage Day (2002), a South African public holiday.
This trend towards only having computers on during office hours can be seen in other subnets too. An interesting observation is that it tends to be more pronounced in those subnets serving departments and divisions that are less computer-centric. At the extreme ends of this observation are the subnets serving the Computer Science department, and the subnet serving the University's Sports Administration, Estates division, and Grounds and Gardens division. The Computer Science department tends to have more computers on the network outside of office hours than within, while the Estates division has almost no computers on outside of normal office hours.
One of the more interesting problems encountered while trying to design a system that generated comparative graphs for various subnets was that of choosing colours for the various lines on the graph. In theory, there could be up to thirty-two lines on each graph. In order to display each of these lines clearly, the colours assigned to each of the lines need to contrast as much as possible. The algorithm used to select these contrasting colours is described below:
Using an additive colour wheel, thirty-two colours were selected from even spaced intervals around the wheel. These colours included the three primary colours (red, green and blue — RGB), as well as a range of intermediate colours between each of the primaries. Each of the colours chosen had approximately the same level of brightness.
These colours were transcribed into an array using their hexadecimal RGB notation. The arrangement within this array mimicked each colour's relative positions to each other on the additive colour wheel. The start and end position were essentially arbitrary but, for aesthetic reasons, they were chosen so that blue was in the middle of the array.
When a number n of colours had to be chosen for use on a particular graph, a step value s was calculated using the formula s = 32 / (n * 2). Starting at the colour whose array index was s, n colours were chosen from the array in such a way that their array index was s greater than that of the previous colour.
This resulted in a set of opposing colours that are evenly spaced around the additive colour wheel, giving us a range of easily distinguishable colours for the graph.