Once I published the graph on the previous blog entry I more or less immediately realised it was wrong – it’s not that the curves are wrong: it’s that they are plotted on different scales.
The red line plots lifetime curve using the working set of the process based on pages accessed in a certain time frame – this gives an average working set size () which is plotted along the x axis.
The blue line is the lifetime curve with a maximum working set of a fixed size (ie it is a simple LRU type page replacement simulation). But it is not scaled correctly against the x axis. Something of a waste of 31 hours of wallclock time!
Happily my 12 core box is now online and so I should have a replot done shortly – my best guess is that it may not change too much, things will be interesting if it does.