Download Crawford-S-dissertat.. - Department of Computer Science
Transcript
Graphing appearance Preliminary feedback from Fluidata noted that whilst the stacked graphs provided a very detailed breakdown of application usage within a group or by a specific host, they often became cluttered and hard to visualise. With over twenty different data series on some graphs, the colours used to represent each often overlapped or blurred in to one, making visualising the data harder still. The first solution to this problem, discussed at length with Fluidata, involved representing only the top ten (or any other suitably small number) data series distinctly on the graph, and then grouping the rest together. This led to graphs such as the following appearing (note the final data series in the table is labelled “Others”). Figure 24 - Traffic graph with minority traffic grouped together However, despite this significant improvement, problems were still experienced. Fluidata noted that when a few pixels on the graph could represent tens or hundreds of kilobytes per seconds, it made it very hard to distinguish between data series. This led to the suggestion of zoom functionality. Research conducted into the possibility of zooming into RRDTool generated graphs soon found a GPL-licensed script known as “Bonsai” that performed precisely this purpose (Steffen, 2004). However, testing of the script discovered a flaw – it only allowed for horizontal zooming (i.e. one could zoom in to a time span, but could not zoom to 50KB/s to 100KB/s traffic rate). The script was modified to support vertical zooming as well, and this modified version is included in the source code in the appendix. 57