Sorry -- I left out the K=1 expressions from that paper. They're a little more mundane (for Stratus) but of obvious interest here. These are the most likely to be interesting to us. Consider, for example, \lambda = 0.5 failures/year and \mu = 52 repairs/year for the kind of machines we're talking about (as opposed to Stratus machines). The resulting MTTF would be in years. - Carl ==================================================================== N = 2 ; K = 1 \begin{equation} \frac { 3\lambda + 1\mu }{ 2\lambda^2 } \end{equation} N = 3 ; K = 1 \begin{equation} \frac { 11\lambda^2 + 4\lambda\mu + 1\mu^2 }{ 6\lambda^3 } \end{equation} N = 4 ; K = 1 \begin{equation} \frac { 50\lambda^3 + 18\lambda^2\mu + 5\lambda\mu^2 + 1\mu^3 }{ 24\lambda^4 } \end{equation}