Visualizing Disk Space

INFO VISUALIZATION: Filelight is an interesting KDE program for visualizing the size of a given directory. Check out this example image:







The first ring around the "doughnut hole" is divided into color coded sections that represent the first subdirectories and their size relative to one another. So, in this example, the entire ring represents 31 megabytes. The next ring out represents the size of subsequent subdirectories and so on.



Overall I like this idea, I wish I could use this on my windows box to free up some space. It certainly beats what I do now -- search for all files over 100 Megabytes and delete anything I don't need anymore. This system of viewing the size of directories rather then files could be more useful for a lot of projects that have a deep file structure.



But the radial nature poses some problems. The biggest issue here is that smaller subdirectories appear much larger then parent directories -- which actually should be represented with larger shapes. This is just the nature of the program since each consecutive ring, although much larger in area, represents the same size. In this case, each complete ring is 31 Megabytes, even though the last ring is several times larger.



Also, I find it a bit counterintuitive that the size of files in any particular directory are not represented (or represented as white space) rather then an extension of their respective directory. This makes it more useful in a context where only the directory size is informative.



And the circular nature brings up the same problems as does the pie-chart. (Ahhh, the pie chart. So loved by MBAs; so hated by everyone else.) Speaking of pie-charts, here's one of my favorite Tufte quotes:

Pie charts make your viewers stupid; pie charts also say that you don't know anything about data. Use a table.

You can view a nice short discussion on "Ask Tufte" about pie-charts' failings here. Also, check out a thread that suggests the use of Nightingale Rose charts (which are interesting, but still less useful then a normalized histogram IMHO).