Today was a good day. A really, really good one. That's why I'm writing about it. I mean, not only did my mind come up with one heck of a theory, but I also was provided living proof for it just a couple minutes later.

My new theory goes as follows: Replace all known measurement units with percentages, and noone will notice.

This theory is backed up indirectly by work done by famous people like Einstein (Everything is relative) and Fibonacci (Mathematics is the language of the nature) in that every graph done in percentages will only display relative numbers, and it will become natural to work with comparisons to other data.

The proof that was delivered for my theory was a set of graphs showing some sort of data gathered from a network interface in a provider's router. There was one graph showing bars representing the total traffic per day, which was labeled in megabytes, and although it probably used the old factor 1024 for kilo, it was one the only usable graphs in the whole statistics document. The other one was the CPU load. But that's the one the trouble starts with.

The other graphs showed non-sense data with scales from 0% up to 0%, or, when lucky, 0% to 100%. And no indicator whether that's bits or cycles or apples or cakes or kittens. There was a graph labeled input load which showed a nice graph that was maxed out for the last two weeks, and my colleague immediately jumped to the conclusion that this had to be the peak line utilization graph, totally neglecting my hints that the total traffic graph showed hardly any bytes on those days, and that it was the exact same curve as the CPU utilization. And even peaks have to have some sort of timespan, otherwise the graph would be showing a 100% all the time, as it'll be maxed out as soon as there's one single data packet going over it. Surely, for a microscopicly short timespan, but practically still maxed out at that very moment.

So, what I'm actually trying to say here, is that there hardly such thing as a percent axis. There's got to be some sort of measurement unit for a graph to make sense. If it's the line utilization, note down the line capacity, for god's sake. There's always enough room for a 2MBits, or whatever there's installed on site. But sure, for CPU load, this is CPU cycles used per second, with cycles being measured in Hertz, and Hertz being one per second, basically boiling down to an amount of ones, with the X-axis showing the timeframe. However, even then, you'd still have the maximum number of ones in that given timeframe.

I guess the moral of that sory is, that you should always sanity-check the graphs that you're publishing, as I have done with the graphs on this site. And the check result is: 0%.