We use a structural operational semantics which drives us in inferring quan
titative measures on system evolution. The transitions of the system are la
beled and we assign rates to them by only looking at these labels. The rate
s reflect the possibly distributed architecture on which applications run.
We then map transition systems to Markov chains, and performance evaluation
is carried out using standard tools. As a working example, we compare the
performance of a conventional uniprocessor with a prefetch pipeline machine
. We also consider two case studies from the literature involving mobile co
mputation to show that our framework is feasible.