The typical performance figures of muskel are shown in the picture above. Coarse grain programs (i.e. those where remote computations take a substantially higher time than the time required to forward input data and gather output data) scale almost perfectly (grain 170 in the picture), while finer grain programs scale worse.

This is quite good result, taking into account the amount of (programming) work needed to run muskel programs (see the code on the right figure).