ahh graphics... they look pretty but one might ask how useful they are. at the same time when I use collectl to gather the data and colplot to look at the results that comes to 86400 samples for a few hundred counters, not something you want to look at line by line. Graphics for this is the way to go to zoom into a problem area BUT then you better look at the raw numbers because you'll always pick up subtleties you might not otherwise see.
I was careful to mention the number of samples and colplot (which actually uses gnuplot) because many people might argue 10 sampling is far too ofter and generates too much system load. In a word, they are wrong. I'll bet there are millions of system admins happily collecting worthless data ever 10 minutes with sar and don't even know it.
THen there's graphics. How many people rely on tools like ganglia? How many of you know the plots are not accurate? We can continue that discussion later if you like.
-mark
