Variability Model of the Linux Kernel

Lack of realistic benchmarks hinders efficient design and evaluation
of analysis techniques for feature models. We extract a variability
model from the code base of the Linux kernel, obtaining a model larger
by an order of magnitude than the largest publicly available feature
model so far. We analyze properties of this model, compare it with
previously available benchmarks, and emphasize the differences from
published academic examples. As a result, we broaden our understanding
of what a feature model is, hopefully challenging tool designers by
providing an interesting benchmark, giving input to design of random
model generators, and last but not least, inspiring designers of
variability modeling languages.