We have seen earlier that the consensus problem cannot be solved in an asynchronous system. We have, however, also seen that there exists randomized algorithms solving the problem in an asynchronous system. Furthermore, we have shown as a homework that mild synchronicity suffices to make the problem solvable without randomness. In this lecture we will study another approach to make the consensus problem solvable in an asynchronous system, namely failure detectors . In what follows we will consider crash failures.