it is common to graph an item across multiple hosts (like a cluster) to see whether things are more or less evenly distributed.
in a large environment, the graph is very crowded, and the legend is even worse.

it might be useful to be able to configure a graph to show average across a host group, and then a configurable number of outliers. so instead of seeing 50 values (with a legend like a small story), we would see avg + 3 outliers - and see the values of those in the legend.

it would be useful to configure methods of determining the outliers.