[Planetlab-users] challenge

[Larry Peterson]

A challenge for PlanetLab users... How does one quantify how
big PlanetLab needs to be. As PlanetLab went from 100 to 200
to 300 sites, how much improvement in fidelity did we see in
our experiments? (And how do we measure fidelity, anyway?) Is
there value in it being bigger, and if so, how much bigger?

Bigger might imply having nodes in close proximity to more
users, which potentially brings more workload. How much
real-user workload do you need to believe your results. Can
we translate workload into a measure of experimental fidelity?

There is a related issue of site selection. On this point, there
may be ways to measure how representative of the Internet PlanetLab
is (or could be), and it has been argued that PlanetLab is currently
too GREN-heavy, but in what way does that impact our results?

Larry